tech

2018-09-17 · open the original ↗

• Windows 10 now ships in six-monthly releases instead of big bangs. Rey, lead trainer at Brainstorm, says Windows 10 was finalized in July 2015 and now updates every six months, the coming one being the Windows 10 October 2018 Update, called Redstone 5 in beta; he says this lets Microsoft respond to features users request through UserVoice. Fall releases are supported for 30 months for Enterprise, Education and, he believes, Pro, while spring releases are smaller and businesses tend to standardize on the fall one.
• Businesses control their rollout; home users just receive updates. Rey says a company's IT decides when a new version installs, helped by Windows Insiders and Project Olympia, which provides an Azure Active Directory at olympia.windows.com into which a tester can join a device. Home machines get updates automatically, though Microsoft has added controls, after feedback, over when the download and reboot happen.
• A cloud clipboard will sync clips across devices. Rey says Windows key+V opens a clipboard that holds more than the last item and synchronizes across machines, so something copied at home can be pasted at work. He thinks it is in the preview for spring 2019 but was unsure whether it is in Redstone 5; Giard gives his own use case, moving screenshots between computers for his blog.
• The Snipping Tool is being replaced by Snip & Sketch. Rey says the new tool, first called Screen Sketch, works much the same but supports pen and touch, and he understands the Snipping Tool's delay feature is coming to it. The Snipping Tool remains for now with a prominent notice that it will go away.
• Focus assist blocks notifications for a set period. Rey says it suppresses Action Center notifications and pop-ups such as new Outlook mail for, say, two or three hours, while chosen priority contacts such as a manager still break through. It sits in the Action Center.

2018-09-07 · open the original ↗

• A storage account holds blobs, queues, tables and files under one unique name. Giard creates it at portal.azure.com from "Create a resource" > Storage > Storage account. The name must be unique because it becomes a URL under core.windows.net, and users reach the account through a REST API at that URL. Files are built on top of blobs and emulate a file share for applications.
• The replication choice decides where the extra copies of your data live. Azure always stores several copies. Locally redundant keeps them in one data center on different fault domains, so they share no rack or power supply but sit within about a hundred yards of each other. Giard says paying a little more for geo-redundant storage keeps a copy in another region, such as Texas for a Chicago data center. A third option also makes that copy readable by your users.
• The rest of the creation form is mostly defaults and optional extras. Giard picks a general-purpose account in East US and the standard performance tier rather than premium ("I'm a cheapskate"). The form also sets secure transfer, the subscription and a resource group, which he describes as a logical grouping of resources managed together. Optional settings put the account on a virtual network for secure access from VMs on it, and enable Data Lake storage.
• Automation options export the configuration for repeatable creation. Before creating, the automation options offer an ARM template with its parameters, plus code in CLI, PowerShell, .NET and Ruby. Giard thinks it downloads as a zip file, which can recreate exactly the same storage account.
• The created account's blades manage containers, tables and queues directly. Creation took about a minute. Under blobs he creates a container (the name must be lowercase, a rule he says he always forgets) and can upload files to it. Tables take an access policy, and both tables and queues show a REST endpoint. Further blades cover security, scalability, CORS and encryption, each linking to documentation.

2018-09-05 · open the original ↗

• Define the problem narrowly enough to know if it's solvable. Seibel's tests: state it in one sentence, ideally have had it yourself, and pick who you can serve first. His case is Poppy, an on-demand babysitting company: narrowed to infants, the problem may be unsolvable, because sitters skilled enough for parents to trust tend to hold well-paid full-time nanny jobs.
• Frequent, intense problems make businesses, and charging tests them. He notes car-buying sites keep failing for shoppers, who buy about once every seven years, and are really built for dealers with daily quotas. He argues starting with a price almost always beats starting free, since free attracts users without the problem whose feedback leads you astray.
• Sell a bad MVP to desperate users, not friends or investors. An MVP is bad by definition, so find the customers who will go out of business without you. A friend he names as Steve said he could finally delete Socialcam once it sold, and Justin fired an abusive Exec customer after four refunds. Twitch came from finally talking to gamers, about 20% of Justin.tv traffic for years.
• Event-based metrics, specified with each feature, are a prerequisite. Seibel says Google Analytics is the wrong primary tool; use an event-based product such as Mixpanel, Amplitude or Heap. Start with five to ten stats (for Instagram: open, create account, take photo, apply effect, share), name them for future readers, and put the expected metric changes in the spec.
• A disciplined two-week cycle replaced Justin.tv's three-month argument spiral. He describes his process: one KPI (revenue, or daily active users if never charging), an open brainstorm, easy/medium/hard sizing, a written spec, and a single meeting per cycle. He calls changing customer or problem a pivot, which should be rare, and expects iterating toward a solution to take about two years, not two months.

2018-08-28 · open the original ↗

• Encoding a video does not by itself give it captions. Giard creates the Media Service from Create a Resource with a new resource group and a storage account for the media, uploads a video on the Assets blade, and encodes it with the defaults. Encoding runs as a job that moves through scheduled, queued, processing and finished; its length depends on the video's length and size, and the result is a second asset with no captions.
• An Analyze job on the original video generates the caption files. By default it writes captions in three formats, all checked; Giard thinks VTT is the most popular and uses it. The same job can also produce index files for SQL Server full-text search and a keyword file; he keeps the defaults with English as the language. Analysis is also queued, which Giard explains as Azure using resources efficiently for a one-time, non-critical process, and it yields a third, indexed asset.
• Auto-generated captions need checking, and the VTT is plain text. To get a download URL, the indexed asset must first be published with a progressive locator; Giard then downloads the file with curl (curl -o poe.vtt <url>) because his browser could not work out its MIME type. The captions of his reading of Poe's "Annabel Lee" contained mistakes, which he corrects by hand in Notepad before uploading the file as captions to the encoded asset.
• Streaming requires publishing and a running streaming endpoint. After publishing the encoded video, the portal shows the streaming endpoint in red until it is started; Giard notes that with no restrictions added, anyone can view it, and that restrictions can be added later.
• Azure Media Player shows captions only after you add a track. Giard pastes the playback URL, without the http: but keeping the //, into the Azure Media Player demo page; by default there is no captions menu. Under the advanced options he adds a track of type captions or subtitles, with a label (several caption sets are allowed), a language, and a URL at the same location as the .ism manifest but ending in the VTT file's name, after which an English captions menu appears. He says step-by-step instructions are in an August 2018 post on his blog.

2018-08-28 · open the original ↗

• Live streaming runs through a channel on the live streaming blade. Custom create lets you set the encoding type, name, description, auto-start, streaming protocol and IP restrictions; quick create takes the defaults, which gave Giard pass-through encoding and RTMP ingest. He warns that you are charged while a channel is running, so it is worth being diligent about stopping it.
• The channel's ingest URL is where the streaming software points. The channel blade lists a primary and a secondary ingest URL, and either works. The channel came with a live event named "default", already started and published but not streaming until a source is sent to it.
• Separate streaming software sends the stream to the channel. Giard uses Telestream's Wirecast, priced at $449 to $949, on its free trial, which stamps a watermark about every 60 seconds, fine for testing but not for production, in his view; install needed one restart. In Output Settings he chooses an RTMP server as the destination, pastes in the ingest URL, adds a screen-capture source (layers can add a camera or branding), moves the preview to live, and starts the RTMP server.
• Viewing needs the streaming endpoint running, and "live" lags. The stream does not play until the streaming endpoint is started; then Watch in the portal shows it. Giard notes it runs a few seconds behind, since there is always latency going out to the web and back.
• Anyone can watch through Azure Media Player. He pastes the channel's .ism/manifest URL, without the http: prefix, into ampdemo.azureedge.net to share the stream with people outside the portal.

2018-08-28 · open the original ↗

• A media service comes with storage and a streaming endpoint. Giard creates it from Create a resource, under Mobile, giving it a resource group, a region (he advises one close to you or your viewers) and a new storage account, whose name has to be unique. Deployment took under a minute and created the resource group, storage account, media service and streaming endpoint.
• An uploaded video cannot play until it is encoded. The encode dialog offers several formats; Giard keeps the default MP4 and job names. The job goes from queued to processing to finished and encodes into several formats for adaptive streaming, leaving the original and an encoded asset side by side.
• The encoded asset must be published and streaming started. Play stays disabled until the asset is published and the streaming service is started; a checkbox on the publish step starts it too. In his demo this took about three minutes, after which the portal's own player played the clip.
• Sharing outside the portal needs a generic web player. Because viewers may not have Azure access, Giard uses Azure Media Player: he pastes the playback URL, which must contain ism/manifest, minus http:, into the free player at ampdemo.azureedge.net. The player then gives a link to send to people, and code to embed the player in a web page.

2018-08-09 · open the original ↗

• Chocolatey installs Windows software and its dependencies in one line. Giard calls it one of his favorite tools: a free package manager, Windows-only, installed by pasting the command from chocolatey.org into a PowerShell or cmd window, probably with administrator rights.
• choco install takes the package name from the site's package list. Running as administrator, he installs Visual Studio Code with the command shown on its package page. Without -y Chocolatey asks before running each script (twice for this package); -y answers yes to all of them.
• A batch file of installs rebuilds a fresh machine unattended. When he reformats and reinstalls Windows, Giard runs a batch file with one choco install <package> -y line per tool, such as Node.js, an FTP client and a PDF reader, and comes back a couple of hours later to find it all installed.
• The result is the same as a manual install. Command-line tools need a new command window before they are recognized; after that code . launches Visual Studio Code, and it also shows up in the Start menu, as if downloaded and installed by hand with its prerequisites.
• choco uninstall removes the app and its dependencies. Uninstalling Visual Studio Code with -y took it out of the Start menu, and code . was no longer recognized in a new window.

2018-08-04 · open the original ↗

• The Castaways network was a LAN-only Ubiquiti mesh for cameras. Brandon Yarborough, Crosstalk's wireless ISP expert, built it for an ABC survival show (premiering August 7) across seven miles of South Pacific coastline: 100-plus mesh locations, NanoStation backhauls, Rocket M5s with omnis, all on non-AC 5 GHz, static IPs, no DHCP or QoS. Cameras needed only 2–4 Mbps each against an expected 30–40 Mbps mesh; the job was privately funded, with no Ubiquiti involvement.
• Shaded solar, not undersized panels, caused the outages. Brandon sized the 150 W panels and 30 Ah sealed lead-acid batteries for 12 hours of full sun, based on his New Mexico tests, but the crews had hidden gear in trees and mangroves, leaving maybe 20% of rated output. Fixes were moving panels, adding panels at key relays, and swapping generator-charged batteries every few days; next time he would fit 250 W (or two 150 W) panels and twin batteries.
• Tides defeated a directional antenna on the anchored relay raft. Even with three anchors the pontoon swung too far, so it got two omnis on different frequencies. A planned airFiber link of just over 30 miles from an internet-connected island was dropped: towers tall enough to clear earth curvature and the Fresnel zone weren't feasible, and Brandon says tides can cut throughput up to 50% when the zone's lower lobe is blocked.
• The host and Brandon rate UniFi as SMB gear, not enterprise. They cite no switch port security and VPN that needs JSON edits to run site-to-site and client VPN together, which Crosstalk refuses to do, recommending an EdgeRouter X (about $50) instead. For hotels of 500-plus rooms the host would use Ruckus, not UniFi. They also advise against VoIP over wireless links and against UniFi's zero handoff.
• Budget a two-to-three-year access point refresh cycle. The host says moving client sites from a 5.6 to a 5.8 UniFi controller left older single-band UAP Outdoor units end-of-life: still passing traffic but impossible to update or provision, within two to three years of install. He recommends phasing out older equipment on that cycle rather than replacing everything at once.

2018-07-29 · open the original ↗

• Azure supports Linux virtual machines extensively, despite the Microsoft name. Giard says that the last time he read, about 30 to 40% of the virtual machines in Azure were Linux. Under Compute in the portal he names Red Hat Enterprise Linux 7.2, Ubuntu 17.10, CentOS and SUSE. Some images are managed by Microsoft and some by third parties, so he tells viewers to check the licensing on each.
• A Linux VM is created the same way as a Windows one. He picks an Ubuntu Server 14.04 LTS image managed by Canonical. He chooses a regular hard drive or a solid-state drive, then public-key or password authentication (he takes password because "passwords are easier"). He creates a new resource group and picks South Central US for low latency from Chicago. The size he selects has 2 CPUs, 8 GB RAM and 4 data disks, at about $81 a month if left running for 30 days straight.
• Optional settings cover redundancy, networking, extensions and cost-saving shutdowns. He points out availability sets for fault tolerance, a shared virtual network or subnet, and extensions such as antivirus and backup, but uses none of them. An automatic daily shutdown at a set time and time zone, with an optional email notice, saves money. Downloadable templates let the same machine be created again from PowerShell, the CLI or .NET.
• Remote access requires opening the inbound SSH port. This setting is required. He opens 22, the default SSH port, and accepts the portal's warning that exposing a port to the Internet is a potential security risk. Deployment took about five minutes.
• Linux connects through an SSH command, not a Remote Desktop file. Clicking Connect on the overview page gives the command ssh followed by user@IP address, where a Windows VM gives a downloadable RDP file. He pastes the command into a command prompt, confirms the prompt about the machine having no certificate, and enters the admin password. From there everything happens at the command line, such as installing software and launching services.

2018-07-28 · open the original ↗

• Prosthetics force a choice between affordable and functional. The team says cosmetic prosthetics are cheap but cannot grasp objects properly, while bionic arms cost tens of thousands of dollars because their myoelectric, neuromuscular interfacing is programmed for each individual and so cannot scale. Their aim is a product amputees can be handed and learn to use themselves, with the arm "learning along the way".
• Computer vision chooses the grip instead of neuromuscular sensing. The arm is 3D printed; a camera in the palm feeds frames to an onboard computer, which detects the object and picks the appropriate grasp, and a muscle sensor worn on the arm triggers the grab on a flex and releases it when the muscle relaxes. In the on-camera demo the first attempt narrowly missed the object, though the host notes it used the right grip, and the second succeeded.
• They moved recognition onto the device so it works offline. The team says they started with Cognitive Services image recognition but hit problems with objects' differing shapes and sizes, so they moved to a Custom Vision model deployed through IoT Edge; the model runs locally and the arm needs no internet connection.
• The project grew from a hackathon entry to world champion. It began in January at a University of Toronto hackathon, where it won Microsoft's API award and the host urged them to enter the Imagine Cup; they won the Canadian national finals in Vancouver and were announced world champions, representing Canada, the day before recording in Redmond. Choudhry, who studies robotics, was interning in California with the hardware while Khan, a computer science student, wrote code in Toronto and sent it to him to deploy and test.
• They intend to turn SmartArm into a business. Choudhry says they have meetings lined up with a Canadian federal ministry for innovation and economic development to plan the launch, and that everyone they have told says it fills a gap; he returns to Canada in September. The prize includes a mentorship with Microsoft CEO Satya Nadella, cash and Azure hours.

2018-07-28 · open the original ↗

• Azure VMs are portable between the cloud and on-premises. Giard describes a virtual machine as a file that emulates a physical computer, with its memory, disk, software and data. Because Azure uses the same hypervisor technology (Hyper-V) as on-premises virtual machines, he says they move between the cloud and a local machine or network "with relative ease".
• The image gallery offers operating systems and pre-installed software. Under Compute in the portal there are versions of Windows Server, Linux distributions such as Red Hat and Ubuntu, and images with software already installed, such as SQL Server. He picks Windows Server 2016 Datacenter. He then sets a name, a regular or solid-state drive, admin credentials (the password must meet rules shown by hovering the warning icon), a subscription and a new resource group.
• Pick a region for latency; bring or rent a Windows license. He says to place the VM near you, your users and any Azure or web resources it talks to, and picks East US. If you already own a Windows license you can use it. Otherwise Microsoft rents you one and the hourly cost goes up a little.
• You pay only while the machine is running. The listed monthly cost assumes running 24 hours a day for 30 days. Starred sizes are the ones available in the chosen region. The summary page showed about 37.6 cents for 60 minutes, billed by the minute. An auto-shutdown, for example every day at 7 p.m. Central, avoids charges for the hours after that. He also notes availability sets, a shared virtual network or subnet, extensions such as antivirus, and downloadable templates (PowerShell, CLI, .NET) to automate creating the same VM.
• Remote Desktop access requires opening its inbound port. This setting is required. He opens 3389, the default Remote Desktop port, despite the warning about Internet exposure. After about five minutes of deployment, the overview page shows the status, region and public IP, with buttons to reset, stop or connect. Connect downloads an RDP file that opens in the Remote Desktop app. He accepts the warning about the missing certificate because he just created the machine, and notes that the first login is slower while it sets up a profile.

2018-07-05 · open the original ↗

• 4G LTE routers suit backup, no-wireline sites and vehicles. David, presenting for Crosstalk, says failover still leaves a blip of about a minute but cuts overall downtime, and LTE can be the primary link where only DSL is available, as he sees often in central Pennsylvania. Ambulances, police departments and fleets deploy them for tracking and staff internet. He chose Peplink because he had them in stock and uses them a lot, naming Cradlepoint and MikroTik as alternatives.
• The MAX BR1 Mini is a near-standalone router for vehicles or small homes. Unlike the Enterprise it has built-in Wi-Fi, plus a 12-volt industrial power port, two SIM slots, SMA connectors for GPS, MIMO cellular and Wi-Fi, and a second LAN port usable as a WAN for a one-time fee of about $99. He believes it can also use known Wi-Fi as a WAN to offload the 4G link.
• External omnidirectional antennas mostly just move the signal outside. David says an omni antenna on the building adds little gain beyond getting past the walls, whereas a directional antenna aimed at a tower can materially improve signal; both attach to the router's SMA connectors over coax.
• The MAX BR1 Enterprise targets networks that already have Wi-Fi. It runs LTE-Advanced, pulling up to 300 Mbps down and 50 Mbps up if carrier and plan support it, and has no Wi-Fi, a USB port, a LAN/WAN-switchable port once unlocked, and a 12 V/48 V terminal block. It has more processing power and faster VPN/SpeedFusion; both models do IP passthrough, avoiding the double NAT that breaks VoIP.
• SpeedFusion makes cellular failover seamless where ordinary failover drops. Normal failover blips for 10–30 seconds depending on settings; SpeedFusion bonds VPN tunnels over each WAN to a SpeedFusion Hub (an ISO run on AWS or similar) or another Peplink site, and Peplink's own demo cuts the line mid-VoIP-call unnoticed. Both units also have a web GUI at 192.168.50.1 and InControl 2 cloud management at $30 a year, with serial-number pre-provisioning, GPS tracking and geofence alerts.

2018-06-29 · open the original ↗

• Iterative design clutters systems with near-identical features that become traps. Crockford's example is HTTP's POST, PUT and PATCH: they overlap so much that people waste time researching and arguing over which to use, and he says a single verb ("store" might have been a better one) would avoid all of it.
• Ask which option can be removed, not which is better. He says "which is better" arguments are weak on both sides, so they never settle. On tabs versus spaces, tabs go: spaces are still needed between words, space users prove tabs are dispensable, and the old case for tabs (networks measured in characters per second) is gone now that code is gzipped and minified.
• Real cleanup needs new languages, which the industry and web resist. Removing features from JavaScript would break decades of code. He says C++ "should have been put in the ground a long time ago" yet keeps growing. He also says Mozilla once put Python in its browser but never switched it on, rather than prove it free of the security problems already fixed for JavaScript.
• Lookalike syntax misleads, so learn JavaScript itself. Crockford says Java and C# programmers moving to JavaScript keep writing C#, and TypeScript "amplifies the confusion" by looking even more like C#. He recommended CoffeeScript for learning but not for production, and says companies are now stuck with CoffeeScript code. The hard part of the browser is the DOM, "one of the worst APIs ever imagined"; a library in front of it helps, but frameworks have grown complex and sticky, so picking the wrong one is expensive.
• His own early mistake was treating JavaScript as Java. He says Netscape, allied with Sun, renamed LiveScript and positioned it as a subset of Java. He wrote it that way from 2000 until reading the language standard showed him it was "Scheme with C syntax", which led to the good parts. At recording he was still writing a larger book, How JavaScript Works, hoped for that year.

2018-06-29 · open the original ↗

• Their show packs one speaker's idea into a few minutes. On the Hello World Show, each presenter gets about 5 to 10 minutes with a visual component, usually a whiteboard, and the episodes go on YouTube. At recording the show was a year old, having begun at NDC Oslo with mob programming as its first topic, and was on episode 32, partway into its second season.
• Content matters more than gear, and audio was the hardest problem. Shotgun microphones picked up conference noise until Carl Franklin advised the headset mics used on .NET Rocks (DPA, they think). The kit is a Zoom recorder, Bose noise-cancelling headphones and a Canon T3i. Downing says her Samsung Galaxy S9+ now films better than her Nikon D7000 DSLR, and they edit in Adobe Premiere Pro.
• A live panel version became its own conference session. In a one-hour NDC slot styled on a late-night talk show, each panellist gets about seven minutes, then questions from the other panellists and the audience. The first, in London in January, had Troy Hunt, Jon Skeet and Scott Hanselman among the panel and filled the largest room; Richard Campbell and Jimmy Bogard were due for Oslo. The live format was suggested by Justin James.
• She started it because YouTube tutorials never get to the point. Downing drew on six years of theatre and work with Disney, and picked Schneidenback as a partner driven to do more than code. She credits him with the larger share of the work, since he also edits.
• They grow it organically and would stop when interest ends. They say a show needs no money or crew, only genuine interest. A YouTube following takes patience, but their speakers bring audiences of their own. They also say they put each other ahead of the show, which has built trust, and that "vulnerability will set you free": publish imperfect work.

2018-06-29 · open the original ↗

• Companies patch software routinely but never "patch humans". Street means security awareness training: spotting spear phishing, distrusting emailed links and attachments, challenging people who should not be on site, and reporting incidents. A driver who crashes the delivery van gets training and is held responsible, and he argues computers can lose a company far more money.
• Information security has failed by seeking technology fixes for behaviour. He says the field has become "janitorial": it cleans up breaches and re-stages machines instead of acting proactively. Infosec should detect and report, with HR enforcing policy. Writing off a "stupid user" excuses the failure to train users and to press management on why training matters.
• A fixed prize lottery turns employees into an intrusion-detection system. His proposal is $1,000 a quarter for one prize. Anyone who reports a suspicious email or incident, or is seen acting security-consciously, gets an entry, so cost stays constant while participation grows. He tells of a friend's company where an attacker in a compromised account emailed another user, who reported it thinking it was a test, which he says stopped a potential million-dollar loss.
• His in-person teachable moments did not scale, so he now trains defenders. For about eight years he was hired to break into banks, hotels and financial institutions, then came back two minutes later to show staff what they did wrong. Now he and April Wright are building a blue-team course that teaches IT and security staff to run their own security awareness programme year-round.

2018-06-29 · open the original ↗

• Stack Overflow's unfriendly reputation is sometimes deserved. Skeet, the site's highest-reputation user, says closing off-topic or opinion questions is fine, because the site is deliberately built for answerable questions (a Formula One car, not an SUV). Name-calling crosses the line, and so can curtness: brief text comments meant to save time read as hostile, and he admits falling down there and being passive-aggressive himself.
• The goal is a repository of high-quality questions and answers. "Psychic debugging" of an 800-line dump helps one person and no future reader. He wants askers to cut it to something like 40 lines that shows the problem, and treats diagnostic skill as a silver bullet beside empathy. He suggests teaching it first, even making a learner's first program a broken one.
• The system should slow new askers and value questions like answers. He proposes checkboxes for new users (waived after perhaps 10 well-received questions), such as confirming whether code is included. His own questions rarely take under half an hour. Question upvotes earn 5 points against 10 for answers, a legacy of the early days, and he says good questions are now the scarce thing.
• "Grow a thicker skin" is never the right answer. People from underrepresented groups who live with microaggressions may simply leave after one more, shrinking the community's diversity, as a Stack Overflow post by Jay Hanlon acknowledged. Skeet now flags rude comments he once let pass, and argues that if almost everyone is nice, borderline behaviour gets easier to call out.
• He intends to use his top-user standing to push for change. He says people listen to him more than they should, so he will use that "privilege" for good. In one strictly time-boxed hour he wrote a 3,500-word brain dump, and at recording he planned a longer blog post of proposals.

2018-06-28 · open the original ↗

• Click-to-Run replaces MSI so Office can keep pace with Office 365. A small component is downloaded first, then the bits come from Microsoft's cloud or an internal source if an enterprise wants control of versions. Derksen says the MSI build still gets security and bug fixes but no feature updates, leaving it at May 2015 functionality, three years behind at recording.
• Scrub out the old MSI install before migrating. Microsoft supplies a scrub script for this. He says he has seen many problems when the built-in update path is used instead, so he recommends scrubbing the old Office components and doing a fresh Click-to-Run install.
• Click-to-Run licenses per user, which catches companies out. It is part of the Office 365 subscription (E3 or E5) and keyed to Azure AD credentials. Under MSI, 1,000 installs on 900 licences meant quiet non-compliance; now the unlicensed users cannot run Office at all. The upside is that each user can install on five devices, including at home. On-premises single sign-on needs ADFS.
• Some legacy integrations break, so test applications first. Old ways of integrating with Office sometimes fail under Click-to-Run, which he attributes to Microsoft dropping technologies companies should not be using. Most cases can be solved with effort; deploying Office 2016 MSI is the last resort, which he advises against.
• The cloud means constant change, and the first step hurts most. A hospital he is working with is jumping from Office 2010 to 2016, eight years in one go, whereas moving Windows 10 from 1709 to 1803 is a small step. He argues small steps lower risk, but IT has to stop treating migrations as projects that end. He has seen Office 365 in production with nobody yet assigned to manage it.

2018-06-28 · open the original ↗

• Companies will invest heavily in chatbots, so engineers should experiment now. Niculita tells Cluj's product and services engineers to learn bot frameworks so they can build bots into product roadmaps or become clients' vendor of choice, instead of doing CRM, e-commerce and WordPress work. At his employer, cloud communications company 8x8, the call-centre use is deflection: answering the top 5 to 20 issues without a human, mainly to cut time on hold.
• Users live in messaging apps, and new app installs have stalled. Citing Forrester and Gartner, he says brands are dropping mobile apps for chatbots. He says that since about mid-2015 the top four messaging apps have had more users than the top four social networks. He also says the average US smartphone user installed zero net new apps in the previous month, whereas a bot only has to be added, for example from Messenger's bots tab.
• Start on a drag-and-drop platform, then move to a coded framework. Platforms need no engineering background and suit an MVP, for instance to test whether 100 people would use it daily. Frameworks in JavaScript, C#, Java or Python, from Microsoft, Amazon or Facebook's Wit.ai, give full flexibility. He advises plugging into a cloud ecosystem that already handles translation, security, logging, telemetry and CI/CD.
• Speech recognition is solved; working out the user's intent is harder. He puts speech-to-text above 95% accuracy without voice training, given decent audio and no strong accent. Turning a request like a laptop under five pounds with a touchscreen and an eight-hour battery into an intent and attributes is harder, but tools now return a data structure to code against.
• Users forgive a chatbot less than an app, so log every miss. If a bot misunderstands two or three times, he says, the user is gone and hard to win back. He advises logging, or even emailing, each time the bot gets stuck, so you can tell a silly question from a lost sale.

2018-06-28 · open the original ↗

• A story can teach kids why a circuit works, not just how. Satrom, a developer evangelist at IoT company Particle, felt electronics education stopped at making a light blink. In Ed Gets His Power Back, a green LED must light up before sunset and learns along the way what resistors and capacitors do, while the story builds a real circuit. He published it through his side business, Carrot Pants Studios.
• The Kickstarter took more work than writing the book. He wrote the book in three or four days, two years after first sketching the idea. Getting the campaign's message right took longest: they learned many parents are intimidated by electronics, and the video script took longer than the book. It raised about $17,000 from about 250 backers, many of them first-time backers, and the book was delivered in January.
• The sequel moves kids from a breadboard to a custom circuit board. In Ellie Saves the Day, about a blue LED, the kit is a PCB Satrom designed, with headers so components plug in without soldering. He calls breadboards "barely adult-friendly" and wants kids to see what real electronics look like. The book stands alone without the first. At recording the campaign was set to launch on June 26.

2018-06-28 · open the original ↗

• The conference deliberately covers soft skills, never hardcore tech. Eaton says Kalamazoo X targets what developers are typically bad at: interpersonal skills, UI and UX design, marketing and branding. Talks that year included Jennifer Marsman on women in technology and Alan Stevens comparing code to telling a story. Code and product demos are left to the Day of .NET and code camps. The upload dates from 2018, but the host asks about a "Kalamazoo X 2011", which places the recording earlier.
• Speakers are invited, and their topics shaped, instead of an open call. Eaton feared sifting 200 submissions, so he asked friends three or four months ahead. The lineup had five Microsoft MVPs and three Microsoft evangelists, none pitching a technology. He steered one testing talk, whose abstract demanded unit-testing expertise, toward why testing matters.
• A single track gives every attendee the same shared experience. The organisers dropped a five-track plan about three weeks before the first event. Attendees then do not have to choose between speakers, have something in common to discuss at lunch, and sit through talks they would not have picked. Talks are capped at 30 minutes (20 the first year), which he says challenges speakers used to 60 to 90.
• Passion counts for more than attendance. About 65 people came. Eaton says organiser Jim Holmes told him to hold the event again even if only 10 showed up, and Giard says Holmes named it and CodeMash as his two favourite conferences. At recording no decision on a next year had been made, though Eaton could not imagine not doing it.

2018-06-28 · open the original ↗

• Cloud computing means paying only for the capacity you use. Marsman, a Microsoft developer evangelist, says the average data centre runs at about 7% utilisation ("I think") because it is sized for peak load, like a pizza shop on Super Bowl Sunday or MLB.com before the World Series. She also cites green IT, and says routine maintenance, cooling and patching come off the owner's plate. She concedes that being unable to see the servers is the scariest part, likening it to early online banking.
• The platform is a cloud operating system plus optional services. She describes Windows Azure, the base of the Azure Services Platform, as a distributed OS for the cloud, with .NET Services, SQL Services, SharePoint, Dynamics CRM and Live Services usable à la carte, even from your own data centre. At recording it was a CTP, "not even in beta". Pricing was unannounced beyond being based on storage, CPU and probably memory.
• A cloud service looks like an ASP.NET project plus a cloud project. The SDK and Visual Studio tools from azure.com, plus an invitation token, give you web and worker roles. A definition file names the settings, and a configuration file holds values that can be changed while the service runs. Publishing produces a package and the configuration file to upload through a portal. The SDK's command-line scripts allow build automation, and PHP runs via FastCGI from the March CTP.
• Scaling up is one configuration value, which suits small and large firms. The configuration's instance count defaults to one and can be raised at once. She argues this protects a startup from being knocked over by a traffic spike from a high-traffic site linking to it, and spares the upfront cost of a data centre. Large firms can focus on their core business instead of running IT.

2018-06-28 · open the original ↗

• MySQL is now a built-in Azure service, not a marketplace add-on. Giard says it used to come from the Azure Marketplace at extra cost, or you installed it on a virtual machine yourself. Now you go to Create a resource, then Databases, then Azure Database for MySQL.
• Creating the server takes a few settings, each with a choice. You can start with a blank database or restore one from a backup. You set an admin username and a password that must meet the rules shown on screen. You pick a location: Giard says to choose it by where the consuming applications, managing users and other resources are, and he takes East US. He takes the latest MySQL version and the default pricing tier, General Purpose with 2 vCores, which the portal estimates at about $131 for 30 days running 24 hours a day.
• The portal's Cloud Shell connects to the server from the browser. The Cloud Shell icon opens a Bash shell, where mysql -h <hostname> -u <username> -p connects; -p prompts for the password. The server comes with four databases. He creates a new one, makes it the default with use, then creates a five-column table, inserts rows and selects them back.
• Applications connect using the portal's ready-made connection strings. The Connection strings page lists them for several data-access technologies. You fill in the username, password, database name and host name, all of which appear on the Overview tab.

2018-06-28 · open the original ↗

• Azure SQL Database is SQL Server without managing the server. Giard describes it as built on Microsoft SQL Server, with almost all the features of an on-premises install. The server, the machine and the virtual machine underneath are abstracted away.
• Create the logical SQL server first; its name must be globally unique. Giard prefers to create the server on its own, though both can be created at once. The name has to be unique among Azure SQL servers because it becomes a public URL, <name>.database.windows.net. He sets a login, a password that meets the rules shown on screen, a new resource group and East US. He chooses the location by where the users and consuming applications are, to keep latency low, and notes there may be legal limits on where data can be stored.
• The database starts blank, from sample data, or from a backup. Sample data is for learning SQL Server or for building an application against test data. Giard takes a blank database on the server he just created. There are three pricing tiers, Basic, Standard and Premium, each sized further in DTUs, a measure of processing and memory. Their maximum database sizes are 2 GB, 250 GB and 1 TB, and you can set them lower to cut cost. He takes Standard and the default collation.
• Connection strings and management features come ready-made in the portal. Connection strings are given for ADO.NET, JDBC, ODBC and PHP, with the username and password left out. Auditing, alerts, diagnostics and geo-replication are switched on by setting properties. The database can be managed with the free SQL Server Management Studio, Visual Studio, the Cloud Shell in the browser, or PowerShell on a local machine.

2018-06-25 · open the original ↗

• A volunteer-built community WISP aims at neutral, affordable service. Josh Shapiro, CEO and founder of the Community Broadband Project in Los Angeles, says it targets net neutrality, customer privacy and underserved areas stuck on DSL speeds; after about eight months of prep and two to three months of field testing it was in beta with about 50 signups, expecting 50–60 clients roughly two months later. His advice to new WISPs is to befriend other operators.
• Donated sector gear was planned onto a rooftop before testing. Cambium donated all the equipment and sponsored the video. Two ePMP 2000 sectors with smart beamforming antennas on a 17-storey Marina del Rey rooftop were planned in Cambium's LINKPlanner. Cambium's representative said a "90-degree" sector is really 60 degrees at 3 dB roll-off, and stretching it to 120 adds another 3 dB loss. He advised keeping one smart antenna on the shelf and buying more only where it improves a noisy sector.
• The vendor quoted list prices for the whole stack. Cambium's figures were Force 180 $99, Force 200 $129, radomes $19, ePMP 2000 AP about $599, sector antenna $219 and smart beamforming antenna $399. Per sector there is a hard limit of 120 subscribers, but the representative said package mix matters more: typically 10–20 subscribers per AP for 25 Mbps plans, with 30–40 seen.
• The longest links worked, but only marginally. At the Santa Monica Pier, 3.54 miles out and skimming a building, a first speed test gave about 25/9 Mbps. A Force 200 about three miles out link-tested at 74/10 Mbps, and an internet speed test gave 47.49/8.46 Mbps at 24 ms. A parking-structure site 3.5 miles out had an unstable signal, blamed on trees or a building.
• A short throw maxed the radio easily. At 0.3–0.4 miles a Force 200 reached about 65 Mbps down and roughly half that up at −50 dBm. Josh called three miles the worst case and the shorter throw the more common one, and named swaying palm trees as an LA-specific obstacle. The crew said 75/25 down/up frame split is typical, and recommended upgrading subscriber modules before the AP.

2018-06-22 · open the original ↗

• The two new EdgeRouters share the same internals. The ER-4 and ER-6P both have a quad-core 1 GHz CPU, 1 GB DDR3 and 4 GB flash, and Ubiquiti rates both at 3.4 million packets per second. Max power is 13 W for the ER-4 and 16 W for the ER-6P excluding PoE. The host rates both strong for home or small office, possibly medium, though he would pick an EdgeRouter Pro there.
• The larger model is essentially the smaller one plus PoE. The host calls the ER-4 the step up from the EdgeRouter Lite, adding an SFP port, and the ER-6P the bigger brother to the EdgeRouter PoE-5. The ER-6P has five gigabit RJ45 ports plus one SFP (not SFP+), with 24 V passive PoE on all five "from what I understand". The WAN can be the SFP for a fiber handoff.
• Bridging LAN ports costs CPU on these models. Unlike the EdgeRouter X, they have no switch chip, so a bridged LAN is switched in software. The host bridges anyway, arguing that a home or small-office network will be hard-pressed to max the device out.
• Change the factory defaults during the wizard. The router ships at 192.168.1.1 with ubnt/ubnt. The host moves the LAN off the common 192.168.1.x subnet and recommends a new strong-password user instead of ubnt. He resets Ubiquiti's odd default DHCP range, notes DHCP can hand out a UniFi controller address, and upgrades firmware 1.9.8 to 1.10.3 by upload, a roughly two-minute reboot.
• PoE is set per port, in UNMS or the GUI. Choose 24 V (two pair) or 24 V four-pair. The host says most Ubiquiti devices such as NanoStations and cameras want plain two-pair. He warns against enabling it on a port without a known-compatible device, and faults the PoE indicator for sitting on the side rather than at each port.

2018-06-21 · open the original ↗

• Modern JavaScript has learned from other languages and cleaned up old semantics. Subramaniam calls JavaScript "the assembly language of the web", popular before it was a good language. He says anyone who has not looked at it in five or six years will find a very different language, and credits its maintainers with not becoming complacent.
• Prefer const over let and let over var; use rest over arguments. const gives immutability, fewer errors and code that is easier to reason about. The rest parameter (...) documents that a function takes many arguments and, unlike arguments, is a real array with array methods. He says there are "zero reasons" to use arguments now.
• Functional style removes accidental complexity, though it takes more effort to write. Inherent complexity comes from the problem; accidental complexity comes from the chosen solution, such as building an empty collection and looping to fill it instead of calling filter, map and reduce. He says the shift was hard for him after decades of imperative code. Redux benefits too: immutable objects can share their internals, so copies are cheap.
• Destructuring pulls out only the properties a function needs. For example, const {firstName, lastName, email} = person ignores the rest of the object, and it works on function parameters too, so functions carry no unneeded properties.
• The new class syntax hides prototypal inheritance behind Java-like keywords. He likes the syntax (including extends) but calls it dangerous, because the semantics differ from the class-based inheritance of Java, C#, C++, Ruby and Python. He prefers prototypal inheritance as more flexible. His book on this is Rediscovering JavaScript (Pragmatic Programmers).

2018-06-21 · open the original ↗

• Q# exists because programs express quantum computation better than circuits. Azariah, lead for the Q# language on Microsoft's quantum team, says the field's state of the art is drawing circuits and reasoning in mathematics. He calls himself perhaps the only language designer not qualified to program in his own language, because real Q# programs are written by colleagues with physics PhDs.
• Quantum computing turns hard simulation problems on their head. He says molecules such as chlorophyll cannot be computed classically: representing them takes more complex numbers than there are atoms on the planet. Feynman's insight was to use a quantum system itself as the computing device.
• A qubit holds one value, but from an infinite space. On a beach ball, a classical bit is the north or south pole, while a qubit's point can sit anywhere on the surface and be moved by rotations. The way qubits combine gives a natural parallelism. He says the hard part is encoding a problem so that it gets a quantum speedup.
• His team is betting on a qubit that resists decoherence. Qubits lose state through interference with their surroundings, which caps how long they can compute. Others have built short-lived transmon qubits, while Microsoft is pursuing a topologically protected qubit meant to hold state several orders of magnitude longer. Its stated goal at recording was one qubit by the end of that year and a commercially viable quantum computer within five years, with no qubit count specified.
• Simulation lets people learn quantum programming now. A laptop can simulate 20 perfect qubits, a large machine about 30, and Azure larger simulations. The Quantum Development Kit is free, with tooling for Visual Studio and VS Code on Mac and Linux. The libraries are open source on GitHub, where the team accepts pull requests.

2018-06-21 · open the original ↗

• Recent SQL Server releases build security into every edition. Koprowski, an independent consultant, says the security features of SQL Server 2016 and 2017 are in Enterprise, Standard and even Express. They are also in Azure SQL Database and SQL Server on Azure VMs, which matters for GDPR's protection of personal data. He says features ship cloud-first in Azure and reach the on-premises product later.
• Cloud databases are encrypted by default, with a choice of key-holder. In Azure SQL Database, new databases and their storage are encrypted from the start. Keys can be left with Microsoft or kept in HSM-based Azure Key Vault. Transparent encryption dates from SQL Server 2008; for SQL Server in a VM he advises BitLocker on the VHDs, since unencrypted MDF and LDF files are readable by anyone who can reach them.
• Always Encrypted protects chosen columns end to end. From SQL Server 2016, columns such as a birth date or social security number stay encrypted in the database, in transit and inside the application, which must decrypt them. The choice of randomised or deterministic encryption decides which database functionality remains. Dynamic data masking instead shows privileged users the data and plain readers a mask of characters.
• Azure's defaults close common doors. He says new VMs accept no outside connections until firewall and network security group rules are created, and connections to Azure SQL Database are encrypted. Easily guessed account names such as SA and root are prohibited.
• Encryption's performance cost is small in normal operation. He says benchmarks show no more than 4% overhead, helped by faster memory, CPUs and SSDs. CPU use is high when first encrypting millions of rows and during backups that are both encrypted and compressed. The larger cost is administrative: DBAs and developers must move off old releases like SQL Server 2005.

2018-06-13 · open the original ↗

• This bridge pair suits short point-to-point links. The NanoStation AC is a 5 GHz pair for links to an outbuilding, barn, guest house or remote camera. Ubiquiti rates it at 450 Mbps and up to 15 km, but the host would use other gear at that distance and calls it very good under about five miles.
• The full model justifies its premium over the cheaper variant. List prices are $99 versus $64 for the NanoStation AC Loco. The Loco ships without a PoE injector, takes only 24 V passive PoE and has no second Ethernet port. The full model takes 24 V passive or 802.3af and passes PoE through its secondary port, which the host uses to power a G3 camera.
• Setup runs over a built-in management radio. The UNMS phone app, which replaced uMobile, connects to it to set a static IP and new password. The radio then times out, and the host disables it on startup so the SSID stops broadcasting. He sets one unit to access point PtP and the other to station PtP, using WPA2-AES, firmware WA 8.4.2 and output power cut to 0 dB for the short test.
• The host's measured throughput fell far short of the rating. iperf averaged 165 Mbps outdoors. Realigning, changing channels and widths, adjusting power and clearing brush from the Fresnel zone left it at about 170. The built-in test averaged 213 receive and 154 transmit, peaking at 503 and 232.5 Mbps. Indoors five feet apart it was worse, at 93.6–120 Mbps.
• The host can't explain the shortfall but finds the speed adequate. He suspects long non-outdoor-rated cable runs of about 100 ft and 50–60 ft, or his test method, and asked viewers for examples above 300 Mbps. Even so, he says 170 Mbps sustained is plenty for a remote barn camera and some Wi-Fi.

2018-06-05 · open the original ↗

• Databricks makes Spark easier to use on Azure. Giard describes it as built on HDInsight Spark, Microsoft's Hadoop on Azure, with Spark as its general-purpose compute engine. He says Spark "isn't always the easiest thing to use". Databricks adds a layer that makes it simple to deploy jobs, create Jupyter notebooks and auto-scale by setting configuration properties.
• A workspace comes first, then you launch it. He creates the Azure Databricks resource with a name, a resource group, Central US and the cheaper pricing tier, and pins it to the dashboard. It took about two or three minutes. Launch workspace signs him in from cached credentials, and from there he can create notebooks, jobs, clusters, tables and libraries.
• Cluster autoscaling and auto-shutdown control how much you pay. He picks Apache Spark 2.3 and Python 3. With auto-scaling on, which is the default, the cluster adds a worker node when compute runs low, up to a maximum of 8 that he says can be raised to 10. He sets it to shut down after 60 minutes of inactivity. His reason is that you pay for every node for as long as the cluster is up. The cluster took about five minutes to create.
• A Python notebook on the cluster runs pandas and Markdown cells. He attaches a Python notebook to the cluster. He sets a file-name variable to a CSV of forest-fire data, imports pandas, and reads the file into a data frame. df.head shows the first five rows, or a number you pass in, and the last rows can be shown the same way. Because the notebook defaults to Python, a Markdown cell has to start with %md.

2018-06-05 · open the original ↗

• This is a very fast bridge pair that needs no configuration. MikroTik's Wireless Wire works in the 60 GHz band and sells for about $140, a figure the host was unsure of, and is rated up to 1 Gbps and 100 m. The two radios come preconfigured, and plugging them in with two laptops on the same subnet just worked. Brandon Yarborough notes 60 GHz is very high-bandwidth but short-range and heavily line-of-sight dependent.
• The host measured near-gigabit throughput across a short link. It hit 850–860 Mbps over about 75 feet, with iperf running between laptops across the link. With the host standing in front of one radio, throughput fell to about 600 and then 500 Mbps. Trees were not tested.
• The host pitches it for temporary or emergency links. He sees it for cases where running a cable is not worth the effort, though he wouldn't trust its rain resistance outdoors long-term. He was told an extra $40 licence turns one unit into point-to-multipoint for up to three receivers, which could share one building's internet across rooftops.

2018-06-04 · open the original ↗

• Azure Notebooks hosts Jupyter notebooks in the cloud for free. You sign up at notebooks.azure.com. Get Started lists sample notebooks, which Giard recommends for learning Jupyter, Python and machine learning. A notebook mixes formatted text with executable code and its output.
• Notebooks live in libraries, which can be made public. Under Profile, then Libraries, you create a library with a name and a URL slug under your account. Making it public lets anyone browse your notebooks. A new notebook can be uploaded, copied from a URL, or created blank or with a default language: Python, R or F#. Giard picks Python 3.6. Opening a notebook starts a notebook server, the engine that runs the code and formats the text.
• Variables and imports carry across cells, and cells can be Markdown. Giard sets a file-name variable to a forest-fire CSV hosted at UC Irvine and imports pandas. He runs cells with the run button or Shift-Enter. A later cell reads the file and shows the first five rows by default, the last rows, or a count you pass in. Switching a cell's type from code to Markdown makes it render as formatted text.

2018-06-04 · open the original ↗

• Cloud means third-party hosting that is highly scalable and highly reliable. Giard gives this definition because he says marketers have "hijacked" the term. He groups cloud services into infrastructure, platform and software as a service, while noting that not every service fits neatly into one. On reliability, he says Azure writes multiple synced copies of your data, three by default, with one live. The Azure Service Fabric monitors them and fails over to another copy when the live one goes down.
• Each cloud layer abstracts away more of what you must run. The cloud grew out of virtualization, which turns a whole machine into one portable file. Running your own data center means paying for hardware, security, climate control and patching, none of which is your core business. IaaS hosts your virtual machines, but you still patch their operating systems. PaaS also hides the virtual machine: you ask for a database or a web app and just use it. SaaS, such as Gmail, Outlook.com, Office 365 and Salesforce, leaves the customer nothing to run.
• Start with SaaS and move down only when you hit a limit. Giard frames the layers as a trade between less work and more control. He recommends using an existing service if one does what you need. Where it does not, combine PaaS services with your own code, and fall back to your own virtual machine only when something like installed software or a registry setting forces it. He says this tends to cost less.
• Elasticity means paying for capacity only while demand needs it. At an electronics retailer he once worked for, about 80% of revenue came between Thanksgiving and New Year's. Hardware bought to cover that peak sat mostly idle the rest of the year, and averaging the capacity would have lost sales. In the cloud you go from 2 instances to 20 and back by configuration, or by CPU thresholds that add and remove instances. His other examples are ticket on-sales, a sudden spike from a celebrity mention, and data science's short bursts of heavy compute. He credits the cloud for the field's recent growth.
• Azure bills by the minute and spanned 50 regions in May 2018. Prices are quoted per hour at azure.microsoft.com/pricing but charged per minute. They scale with machine size and rise linearly with instance count. As of May 2018 he counts 50 regions in 140 countries, which he says give near-unlimited capacity, low latency near customers or data, and copies across regions for fault tolerance. A status page shows whether a fault is yours or Azure's. In his web-app demo the name must be unique because it becomes <name>.azurewebsites.net. The app service plan sets location and size, and Standard (one core, 1.75 GB) is estimated at about $74 a month.

2018-06-04 · open the original ↗

• ML Studio turns a trained model into a deployable web service. This builds on the census-income model from the previous episode, which predicts whether a person earns over or under $50,000. Of that episode's two algorithms, which scored nearly the same, Giard picks one. Choosing Set Up Web Service, then Predictive Web Service, creates a new experiment with web-service inputs and outputs. After it runs, Deploy Web Service publishes it to Azure.
• Request-response calls POST JSON and can score several people at once. The documentation gives the URL, the headers, and a JSON body holding an array of inputs, so one call can carry multiple people. It also describes the JSON response and each field's type, and gives sample console-app code in C#, Python and R. A test console can fill in the inputs from the 30% of data held back for testing, and returns the actual and predicted income.
• Batch scoring takes a CSV and three separate calls. First you submit a batch job whose body points to an uploaded CSV, then you start it. A 200 response means only that it started, not that it finished. You then poll the status with a GET, which returns not started, running, failed, canceled or finished. Finished is the value 4. Sample code is given for C#, Python and R.

2018-06-03 · open the original ↗

• ML Studio's drag-on shapes replace machine-learning plumbing code. Giard's point is that steps common to most solutions are shapes you drag on and configure, so your time goes to what is unique: getting data, splitting it into training and test sets, and cleaning it. It is free at studio.azureml.net, or you can use a workspace in an Azure account to get Azure's scalability.
• One-line Python scripts and built-in shapes clean the census data. The built-in Adult Census Income Binary Classification dataset labels income as under or over $50,000. Because 90% of rows have the United States as native country, he filters to those with a one-line Python script, cutting 32,561 rows to 29,170. The script's azureml_main function takes up to two data frames. Clean Missing Data, set to remove the entire row, then leaves 27,504.
• Choosing features takes domain knowledge and avoiding redundant columns. He keeps the numeric years-of-education column over the text education column, since keeping both would double-count education. Using Select Columns he keeps age, years of education, marital status, occupation, race, sex, hours per week and income. A second Python script maps marital status from seven values to married or not married, which splits 47% to 53%.
• Hold back test data and use Microsoft's cheat sheet to pick an algorithm. Split Data defaults to 50/50, but he uses 70% for training and 30% for testing, because a model has to be tested on data it has not seen. Microsoft's Azure Machine Learning algorithm cheat sheet points to two-class classification. He rules out the option it marks as built for more than 100 features and picks Two-Class Averaged Perceptron. Train Model needs income set as the column to predict. Score Model and Evaluate Model show about 81% accuracy, against a coin-flip baseline.
• A second algorithm in the same experiment scored almost the same. Two-Class Logistic Regression, trained and scored on the same split, reached 82% against 81.9% for the perceptron. Giard says the next step would be weighing false negatives against false positives if one costs more, or changing the columns or their weights. His verdict is that this is a strong starting point built with about six lines of code.

2018-05-29 · open the original ↗

• A system that does seven things is a blockchain. Reilly, who works in Microsoft's Commercial Software Engineering group, lists them: peer-to-peer networking with no client-server model, consensus on the system's state, messages as transactions ("add a dollar, reduce a dollar"), a state machine built from those, an open protocol others can join, proof of work so no one party can rewrite transactions, and the distributed database that results. He says Ethereum reaches consensus with Paxos, a leader-and-majority model, and calls proof of work "really hard to determine, but really easy to check".
• The system secures itself, so no server has to be trusted. Reilly contrasts this with the client-server model, where users trust a server that "might not follow best practices". Data signed with elliptic curve cryptography can be verified through the public key without ever exposing the private one, and he repeats the claim that reversing an ECC key would take computing power enough to "boil all the water on the earth".
• Ethereum stores logic as well as data, under a fuel limit. Reilly calls it a distributed virtual machine built "for developers by developers", and says its protocol, APIs and tooling keep changing with little documentation or versioning, because it is early. Its contracts are written in Solidity, which he likens to JavaScript or Java, and every transaction must be paid for with "gas" so that no infinite loop can use up every node's resources; he says this restricts what you can build.
• His team built a Minecraft land registry on Ethereum. A contract records ownership of each plot in a 100 by 100 grid, and an Azure function checks it whenever a player on a modded server tries to mine, rejecting players who do not own the land. Reilly says the game calls the contract the way a C# console app would query a database, with "no database layer… no server layer".
• Bitcoin is a blockchain whose stored data is account balances. Reilly describes it as a giant list of wallets keyed by public key, with each transfer written to the chain, so everyone agrees on what everyone holds without a central bank. For learning more he points to the Bitcoin and Ethereum white and yellow papers, which he believes are on GitHub, and to his site timmyreilly.com.

2018-05-29 · open the original ↗

• Drupal is a content management system, a framework and a community. Byron, whom Acquia sponsors to work as Drupal's product manager, community manager and core committer, says it powers about 2% of the internet. As a framework it can run fully decoupled: she cites Princess Cruises running ship manifests on a Drupal back end, and a USB key carrying Drupal for emergency communications.
• Drupal trades setup speed for control, unlike WordPress. Byron says WordPress suits a site that must be up "in five minutes", while Drupal suits "ambitious digital experiences" where the owner wants full control. She likes it for customers who don't know what they want, because a site can grow into e-commerce or new integrations and still look consistent; Giard frames this as the trade-off between work done and control kept.
• Anyone can publish a module, but only vetted ones get security coverage. Byron describes two tiers: anyone who agrees to submit GPL-compatible code can publish, and stable modules that pass checks such as for XSS get a green mark showing the security team covers them. She says this matters because Drupal holds sensitive data, in healthcare and in an online petition system on whitehouse.gov under the previous administration.
• No company owns Drupal; contributors come for their own reasons. Byron counts about 4,500 people on the core, of whom 200–300 are active all the time, and about 35,000 developers building modules and themes. They contribute because a boss chose Drupal, an analyst report or high-profile site drew them in, they have "an itch to scratch", or patching upstream beats a local hack for a customer.
• The stack is PHP, GPL and any database, moving toward JavaScript. Drupal has run since 2001 on PHP with MySQL traditionally, though anything from SQLite to MongoDB or Oracle works, on Linux or Windows, and Byron puts its modules at about 30,000. JavaScript such as React is being added on the client side, not yet the server. She credits Microsoft with work making Drupal run on Azure and with sponsoring zip downloads beside the tarballs that were hard to handle on Windows.

2018-05-29 · open the original ↗

• Modern C++ means working in abstractions, not pointer loops. Gregory, who runs a consultancy in Ontario, says hand-written loops that increment or subtract pointers still compile but should give way to C++11's ranged for and standard library algorithms such as find, which name what they do. Her rule: "the best code is code you don't have to write or test"; rely on the library, and where you cannot, put the code in a named function.
• A standards committee, not one vendor or its creator, governs the language. Gregory contrasts the ISO standards process of national bodies, twice-yearly meetings and papers with C#, where a Microsoft team decides; she says Bjarne is still active but "not the dictator" and the committee sometimes overrules him. Vendors may ship proposals years before they are standardized, as Visual C++ has with coroutines because Microsoft is heavily involved in that proposal.
• C++ pays off when you must control performance, very large or very small. Gregory cites big numerical simulations of finance and weather, and small devices where what counts is instructions per watt, such as programmable jewelry that 12-year-olds were learning C++ to program. She says every robot on Mars runs C++ and calls it the most general-purpose, cross-platform option, though some microcontrollers leave it the only choice.
• Teach C++ from day one, references before pointers. Her "Stop Teaching C" talk objected to intro courses that spend early days on C and teach features in the order they were added. Teaching C++ first gets students further, to templates, exceptions and polymorphism, and she says people from C are horrified at leaving pointers until later "but it really works".
• Her clients stay on Windows because old libraries tie them there. Gregory works mostly in Visual Studio with Visual C++, because clients built around MFC, "well over 20 years" old, cannot port to Unix overnight; she notes Clang and GCC can run from inside Visual Studio. She points beginners to isocpp.org, the site of the standards committee and the C++ Foundation.

2018-05-25 · open the original ↗

• Master basic models before reaching for deep learning. Miller, a senior data scientist and instructor at the Metis boot camp, told his Ideas Conference talk that linear and logistic regression come before TensorFlow and Theano. Without the simple techniques, he says, you cannot understand what deep learning does, why it is built that way, or how to prepare data for it; students still arrive planning computer vision as a first project.
• Data scientists now have to be able to write real code. Miller says that perhaps five years earlier, loading data into scikit-learn was enough to be well paid. The employers Metis works with now want hires who can pull data from their warehouse, clean it the way their team does and write reproducible code, so "the balance has definitely shifted" toward software engineering.
• Data science should follow the scientific method. Getting data into a model is not the end, Miller says: you must first frame the business question and design a metric and a test that shows whether it has been answered. That means a hypothesis, an experiment and a check for a statistically significant result, which he says the field is moving toward but must keep moving toward.
• Newcomers struggle most with open-ended, creative problems. Miller says students from report-building jobs want a checklist, but the brief is "can we optimize this process? Go", often with poor data; he gives "60% of the data is missing" as an example. It is often unclear what to classify, such as whether a group targeted with an ad might buy later rather than whether one person buys.
• Data science is math and a little code, not magic. Miller lists linear algebra, calculus, statistics and some Python or R, all learnable free online, and suggests starting with linear algebra. He puts self-study at perhaps 12 to 24 months against Metis's 12 weeks of nine-to-five instruction, saying his employer's format condenses the same material.

2018-05-23 · open the original ↗

• Which camera wins at night depends on viewing distance. In the host's own overnight test of three variants, the standard UniFi G3 bullet, the G3 with the UVC G3 IR extender, and the G3 Pro, he concluded the extender and the Pro suit subjects further away, while the plain G3 suits close-up views.
• Brighter IR washes out faces close to the camera. On the G3 with the extender, the host was recognizable at about 20–25 feet but within roughly 10 feet his face became "a ghost". The plain G3 barely showed a person at 20–25 feet, yet he was recognizable at about 15 feet, where the extender-equipped camera washed him out.
• The extender-equipped G3 beats the G3 Pro on night-vision value. The host judged their night footage about the same, with the Pro perhaps slightly clearer from its better lens. At about $214 for the G3 plus extender against $299 for the Pro, he found the gap not worth paying for night vision alone; the Pro's advantages are daytime clarity and a wider temperature tolerance.
• The vendor's IR range claims leave the baseline unstated. The extender claims up to 25 m (82 ft), but the host notes the datasheets never say how far the cameras' built-in IR reaches. The extender cost $72.48 on Amazon at recording; fitting it meant forcing out the sun hood with pliers and backing out screws, and it stays dark until "Enable external accessory" is switched on under the camera's Video > Accessories settings in UniFi Video.
• A cheap standalone IR illuminator is the host's proposed next test. He bought one for about $22–24 on Amazon and wants to see whether it outperforms the $72 extender added to a G3 he puts at $140-plus; this is a plan, not a result.

2018-05-16 · open the original ↗

• Feynman's problem, not an application, kicked off the field. Preskill traces the motivation to Feynman's early-1980s argument that simulating quantum nature requires a quantum system rather than a digital one; the idea gained urgency a decade later when Peter Shor showed quantum computers could factor large integers, and skepticism about decoherence was answered by quantum error correction theory developed in the mid-1990s.
• Entanglement hides information from the environment, and that's the trick. Preskill likens it to a quantum book of pages: reading any single page yields only noise, since the content is encoded in correlations across all pages, not in the pages themselves. Quantum error correction protects information the same way, so environmental interference can't detect what it's trying to disturb.
• No hardware approach has won yet. Preskill expects devices with 50 to 100 qubits within a few years, already too complex for classical simulation, but the real limiter is gate error rate — currently near one in a thousand — not qubit count; trapped ions, superconducting circuits, silicon electron spins, and Microsoft's unproven topological qubits are all still competing to scale.
• He sees entanglement, not geometry, as physics' next frontier. Preskill argues space-time's structure is likely emergent from quantum entanglement rather than fundamental, calling this the "entanglement frontier," and predicts that near-term quantum computers, by building small highly-entangled systems no classical computer can simulate, could offer experimental toy models of that emergent geometry within a couple of decades.

2018-05-09 · open the original ↗

• An older mobile platform still pays off for Hunter. He says the return on investment from Windows Phone has been very high, with "hundreds of thousands" of active users a month for the apps he and his wife build. He uses Xamarin for Android and iPhone and says the apps are basically the same apart from following each platform's conventions.
• An app succeeds if users finish it happier than they started. Hunter says the developer's job is to make a complex or mundane task "inconsequential or delightful". If you have to tell users how to use your app, he says, it is probably too complicated: the UI should show what is happening so users think about what they are doing, not how.
• Apps are bloating, chasing wow factor over ease. Hunter says wow wears off after a couple of days and starts getting in the way, and showcase features often go unused. Developers build for themselves, he says, when users are seldom like them: they are busy people for whom the device is an appliance, so a utility app should get the job done and get out of the way.
• Make the simplest entry the default and let detail be added inline. In his finance app, which he would not name, a user can record a receipt as "I spent 50 bucks" or add line items such as $32, $6 and $4 tax that total into one entry, on the same screen. Hunter rejects a settings toggle for this, because it adds steps.
• Split a complex domain into separate apps that share data. Hunter would pair a simple budget app ($300 of spending money, $260 spent) with a detailed account-tracking app, linked in the background so line items match up to the budget. Giard likens this to microservices with a UI, and Hunter agrees.

2018-05-09 · open the original ↗

• Imagine Cup is a worldwide hackathon of student-built projects. Matheney, who worked with Giard on Microsoft's academic team recruiting entrants, calls it "a hackathon of hackathons" for projects built on Azure; the episode was recorded at the US finals in San Francisco on 5 May 2018, the competition's 16th year. Country winners go on to world finals with $100,000 at stake; she first gives the minimum age as 18 and later corrects it to 16, for high school through PhD students.
• A large field of entrants was cut to a dozen US finalists. Matheney says about 4,000 students registered and 221 teams submitted projects, though she later says 121. Six of the 12 advance to the world finals, and some teams first formed at other hackathons, including two from the Vatican's VHacks whose members lived far apart.
• Judging weighted technology use and business case equally. Matheney says about 50% of a score was the use of Microsoft technology and 50% the business plan, polish and user testing, judged on a five-minute pitch and five minutes of questions. The top six win $10,000, $8,000, $6,000, $4,000, $2,000 and $1,000, and a team can win at most one special award for mixed reality, big data or AI, so the most any team can take home is $11,000.
• Many finalists built from personal experience, often in health. Projects included Loro, which adds a 360-degree camera and voice recognition to a wheelchair, a medicine reminder, an asthma-inhaler tracker, and a refugee app that uses the Cognitive Services vision API to match photos of missing family members. Matheney highlights a presenter who told of growing up with a stutter; Giard adds others inspired by a brother born without enough oxygen and a mother who died of a disease.
• The finals included preparation for teams and outreach to children. Matheney says teams had talks, a panel on startups versus big companies, pitch coaching and one or two technical mentors each from 11:30 to 7:30 the day before. While judges deliberated, K–12 children from the Bay Area, including junior members of the National Society of Black Engineers and Girl Scouts, programmed Lego robots. Registration for the next year is at imaginecup.com.

2018-04-29 · open the original ↗

• A wristband sensor turns kids' dribbling into a live team game. Ashley describes a project built during the week of recording for the Seattle Storm's (WNBA) youth camps: kids split into two teams of roughly two to ten, each wearing an IMU wristband, and an app projects each team's dribble count and a timeline so they can see in real time who is winning.
• The sensor reads enough motion to tell what the ball is doing. Ashley says the IMU samples at about 100 Hz, recording g-forces as a 3D acceleration vector and angular velocity of hand and wrist, which he calls a "surprisingly huge amount of information" about the ball as it leaves the hand.
• The model runs on the chip so it works disconnected. The team recorded their own raw dribbles at 100 samples per second, built the model in R and Machine Learning Studio, then converted it to efficient C code in the device firmware; the sensor sends events to a mobile app only from time to time, because sports venues may lack a cloud connection.
• The device code is open source for partners to reuse. Ashley says the sensor kit SDK, which he and others collaborated on, is fully open source and was described in an MSDN Magazine article in April 2018; Giard says he wrote the API linking the device to the web server.
• Basketball is a start; more drills and sports are planned. Ashley's next steps are analysing wrist rotations and supporting more sensors and devices so partners can extend to new sports; he says the project's original push came from snow sports, and Giard suggests baseball or golf swings.

2018-04-22 · open the original ↗

• Windows 8 aims to be one OS from smallest device to largest. Kunk, speaking after the Build conference released the Windows 8 developer preview, says it has a desktop mode essentially unchanged from Windows 7 and a touch-based Metro mode modelled on the Windows Phone 7 live tiles; he contrasts this with Apple's separate desktop and mobile operating systems.
• Desktop developers need to learn nothing new; Metro means real change. He says Metro sits on WinRT, a new base API talking to the kernel in place of Win32, and its apps run sandboxed with restrictions such as no walking the whole file system, which he says is meant to force responsive, safe, phone-like applications.
• Metro offers two UI routes, both on WinRT. C# or VB with XAML, where Silverlight and WPF skills transfer well, or HTML5 and JavaScript as a first-class language for native Windows apps; he stresses that Build's HTML5/JavaScript talk meant Windows applications, not browser ones, and that its appeal is moving existing web code onto Windows.
• Learn HTML5, CSS and JavaScript first for greater reach. Asked where an unprepared developer should spend the next couple of years, Kunk calls the answer obvious: those skills also work on the web, while a Metro app still targets only Windows and will not run on a Mac.

2018-04-05 · open the original ↗

• GDPR binds any company holding EU citizens' data, wherever based. Rialdi, a 15-year developer-security MVP, says the regulation covers how personal data is transmitted (HTTPS rather than HTTP), stored and made available; companies are scared because fines can reach 10 or 20 million euros, and it takes effect in May (of the 2018 recording).
• Nothing in it is new to security experts; it removes procrastination. He says its approach mirrors established practice such as Microsoft's Security Development Lifecycle and OWASP, which he thinks inspired it; what changes is uniformity across Europe, where he says Germany has treated the issue most seriously.
• Start with a threat model; no tool certifies compliance. He calls the SDL threat model central for finding architectural risks, ideally begun at the start of design but for existing systems before May, and recommends Microsoft's free threat modeling tool, while stating there is no tool that can tell you whether you are GDPR-compliant.
• The regulation sets goals and suggestions, not technologies. Rialdi says it asks you to evaluate pseudonymization or encryption and always protect the channel; he adds it is not only a legal document but contains technical prescriptions, so developers should read every point and take advice from security specialists.
• Never store passwords in clear, even for trivial data. Because users reuse passwords, a leak from a site holding only names and emails can open what he calls a cross-vulnerability into, say, a health system; he says it is unclear how EU authorities will treat this case, but a legal contact told him to avoid it absolutely.

2018-04-05 · open the original ↗

• A progressive web app is a website that works offline and installs. Bello, of Southworks and a Microsoft MVP and Google Developer Expert, says new web standards let a site keep working on low connectivity via the service worker and manifest, and let a frequently visited site prompt users to add it to their home screen.
• It spares building separate native apps per platform. He contrasts writing Objective-C for iPhone, Java for Android and HTML/JavaScript for the web with using browser standards that now reach device features such as camera and microphone without installing a native app.
• The basics are only browser, HTML, JavaScript and CSS. No framework is required because these are web standards; he uses VS Code and browser dev tools, with Gulp, Grunt or npm scripts to automate tasks such as generating the service worker, and names Lighthouse (now absorbed into Chrome) for PWA audits and Workbox for service-worker caching.
• Frameworks organize the app, not the PWA part. Bello writes plain JavaScript for some projects and Angular or React for others, using them to structure larger applications; he says the field is new enough that more tooling will appear.

2018-04-05 · open the original ↗

• .NET Conf grew from one Uruguayan event to five countries. Fernandez, CEO of Kaizen Softworks, a Uruguay-based software firm serving US companies, says he and fellow community organizers founded it in Montevideo, ran it there for three years at about 500 attendees, then expanded to Argentina and Colombia, with Chile and Peru added this year.
• The new .NET Conf Global aims beyond conferences. Founded a couple of weeks before the interview, its stated mission is an ecosystem connecting students, professionals and speakers to share knowledge of Microsoft and open-source technologies through conferences, meetups and forums.
• Founders co-run a country's first edition, then step back. Fernandez says the original team actively organizes each new country's debut, including visiting months ahead to pitch sponsors, and becomes a passive collaborator from the second edition once the local team has learned the format.
• South America had lacked community-run .NET events. He says the .NET community there had gone quiet and the only earlier conferences he recalls in Uruguay were Microsoft-run, more than 10 years ago; he suggests US events, long established, need to find their own differentiator.
• The 2018 schedule runs September to November. Speaking in early March 2018, he gives Argentina in September, Uruguay in October and Colombia in November, with Chile and Peru dates still in planning.

2018-04-05 · open the original ↗

• Microservices trade one monolith for many coordinated small services. Bellman says splitting an app into single-purpose services lets each scale independently and be written in a different language (C#, Java, C++), and makes them easier to maintain and test, at the cost of having to coordinate many services.
• Service Fabric handles that coordination, and Azure itself runs on it. He says it does orchestration, service location inside the cluster, scaling and health monitoring; it was born to make SQL Database happen, and Cosmos DB runs on it too, which he offers as evidence it is well tested.
• It runs on Azure, on-premises or other clouds, Windows or Linux. On AWS or your own hardware you run Microsoft's bits but manage the machines yourself; apps can arrive as a packaged existing executable (tooling in Visual Studio 2017), as stateless or stateful services built with the SDK, or as Windows or Docker containers.
• Scaling works at two levels: service instances and cluster hardware. Instances scale on application metrics such as HTTP queue length within the same VMs, while the cluster adds machines on hardware metrics such as CPU above 80%; the value over App Service autoscale is packing containers or VMs to use hardware to its limits.
• His Microsoft project splits AI and API across two clusters. Bellman says a Linux Service Fabric cluster runs custom-built Python AI models (not CNTK or TensorFlow) for improving Microsoft business processes, and a Windows cluster runs the API and moves data from Cosmos DB to SQL, with Service Fabric handling scalability.

2018-04-05 · open the original ↗

• Data science is a process, not a set of technologies. Christopher, director of technical curriculum at Pluralsight who curates its data content and says he is not a practitioner, defines it as something like the scientific method applied to data sets to turn data into action.
• Data science is following computer science's path to commoditization. Christopher notes CS degrees did not exist in the 1960s and the field then demanded one; Giard adds that mainframe costs kept out anyone without a degree until 1980s PCs costing a couple of thousand dollars. Christopher says cheap, on-demand cloud compute and storage now remove the hardware barrier for data science.
• It is happening faster because software lessons carry over. Christopher says an AI system that took him six months to build in school 20 years ago, without enough data to train it, can now be running in an hour, and building AI systems can be automated to let them compete for the most effective.
• The barrier is now knowledge, and its absence is deceptive. He has seen developers in Azure ML's drag-and-drop tool try options until something seems right; without knowing, say, logistic from linear regression, they get something that looks like an answer, which an experienced data scientist could tell will not work.
• Entrants need programming, analytics and business acumen. His most direct route is a degree, possibly online; programmers should learn clustering, classification and neural networks from courses, books and blogs. He says he lacks business acumen himself and agrees with Giard's suggestion of partnering with someone who has it.

2018-04-05 · open the original ↗

• A good chatbot is one users cannot tell from a human. Leung, a software engineer on Microsoft's team formerly called DX (Developer Experience) and now CSE (Commercial Software Engineer), works on "conversation as a platform". He calls the "press one for this" phone menu a bot, but not an intelligent one. Bots like it save the cost of a person routing calls, and he says adding natural language processing is what makes them intelligent.
• LUIS trains a language model from examples instead of hand-written matching. In Microsoft's Language Understanding Intelligent Service, at luis.ai, a developer maps utterances (the things people say, such as "hello", "howdy", "what's up") to an intent (the verb, here a greeting) and marks entities, the useful parts of a phrase. Then they press train. Leung says the alternative is regex or a giant case statement listing every phrasing and common misspelling. The model is exposed as an API endpoint: the bot passes in the query and gets JSON back naming the intent.
• Bot Builder V4 is being developed in the open on GitHub. Leung builds with the Microsoft Bot Framework's Bot Builder. He says V4 "just came out" and has been worked on open source since about two weeks before recording, whereas V3 was not built in the open. Its SDK installs from NuGet for C#/.NET or as an npm package for Node.js, and he invites viewers to file issues and contribute.
• The bot connector lets one deployed bot reach many channels. The bot can run locally, on Azure or on any cloud host. The connector routes it to channels Leung puts at "more than 12" and later "14 or plus", among them Slack, SMS through Twilio, Facebook, a website and Cortana. His argument is that users never download an app: they talk to the bot on platforms they already use.
• Leung is building a voice bot for getting around Chicago. It is a work in progress. It answers questions about CTA buses and trains and walking routes from point A to point B, spoken through Cortana by way of the connector, and LUIS interprets the many ways of asking ("take me here", "I'm hungry, I want to find food"). Giard fills in that some other service must supply the actual departure times, and Leung agrees. For more, Leung points to dev.botframework.com.

2018-04-02 · open the original ↗

• Cloudflare pitches 1.1.1.1 as a privacy-first alternative to ISP DNS. Launched on 1 April 2018 (not a joke, per Cloudflare's blog), it joins resolvers like Google's 8.8.8.8 and Level 3's 4.2.2.2; Cloudflare cites censorship such as Turkey's 2014 Twitter block, and says it wipes query logs within 24 hours and never writes querying IP addresses to disk. The host calls it far more private than an ISP's resolvers but no substitute for a VPN.
• DNSPerf ranked 1.1.1.1 the fastest public resolver at launch. On DNSPerf's worldwide, last-30-days average it led OpenDNS, Quad9 and Google, with Level 3 outside the top ten. The host cautions it had launched the day before with little traffic, and wants to see where it stands in about six months.
• It supports encrypted DNS lookups from day one. Cloudflare launched with DNS over TLS and DNS over HTTPS, open standards the host notes are not yet widely used but expects to see adopted, since the original DNS standards, about 35 years old, ignored security and privacy.
• Changes to Cloudflare-hosted sites resolve instantly, with no TTL wait. Because the resolver is integrated with Cloudflare's network, an IP change to a Cloudflare-hosted site (or one node of a load-balanced group) needs no time-to-live expiry or propagation, which usually takes up to an hour; across millions of sites the host expects this to matter.
• Use 1.1.1.1 primary and 1.0.0.1 secondary wherever DNS is set. The host enters both as custom upstream servers in Pi-hole (not yet built-in options), notes they can go in the DHCP name server fields of a UniFi network, or be set per adapter under IPv4 in Windows 10; an nslookup of slashdot.org confirmed a quick answer from Cloudflare's resolver, though he did not benchmark it.