Custom Extension Springs Specification: Fill & Download for Free

You'll find 2 major differences at Google:ScaleTechnologiesScaleGoogle deals with more data than any other company on the planet. When I was an intern in 2008, I was doing MapReduces over 1 Petabyte tables.1 Petabyte in 2008, back when that amount of storage required a datacenter!They have an army of machines, which you can use to do whatever you want, assuming the jobs have low priority.Google has resources you could only dream about at other companies.TechnologiesGoogle rolls their own for everything. Some examples:Instead of AWS, you configure jobs on Borg.Instead of MySQL or Mongo, you use BigTable.Instead of Jetty or Tomcat, you use the Google Application Framework.Instead of Spring, you use Guice.Instead of JQuery, you use Google Closure.This can be problematic when you leave because almost all the tools you know are specific to Google. No one else is using them.Fortunately tools are easy to learn.Once you've learned all the custom technologies, Google is full of generic web programmers. Some examples of stuff I worked on:Building automation frameworks for Webmaster Tools.Building internal web tools to pull data from Google+.Building backlinks features for Webmaster Tools.Chrome Extensions to help automate Chrome testing.Of course the opportunity exists to build cool stuff. You could be a platforms engineer, optimizing Google's data centers. Or you could spend your days messing with Android or self driving cars.But more likely, you'll add the 1000th tiny feature in Gmail, or you'll build tools the outside world never sees.I went into much more detail in a blog post.photo credit: Pyramids of Giza via photopin (license)

I’m a purist — I refuse to use frameworks I’ve built, unless I have no other option :)But seriously, I do have nuanced / contradictory opinions on this, because I strongly believe both that:Home-grown frameworks are usually not the right long-term technology stack; open / public frameworks have community support and will mature in a way your custom framework will not.You will vastly grow your skills as a developer and architect by building a framework from the ground up.Reconciling the two is a bit of a balance that I don’t have hard-and-fast rules for, but I have general opinions at least.Public frameworksThe fact is, you’re not going to successfully rebuild all the features of Spring, React, Hadoop, or any other widely-adopted software framework. Open Source has eaten the world because these “tried and tested” frameworks:have iterated on the API several times, drawing clear lines between the framework and user-code.are documented, tested and ready to go.have hundreds or thousands of contributors. And contributor doesn’t just mean “writing source code” — it means “finding bugs, giving API suggestions, writing documentation, providing examples”Writing your own home-grown framework might help you move fast at first, but exposes you to risks the second you stop building it:Stagnation: as soon as you stop typing, the framework stops evolving. You aren’t going to passively pick up new features.Bugs/regressions: every point-upgrade of an underlying library is going to break your framework somehow. Do you want to fix it, or do you want Someone Else to fix it?Skills: maybe you understand MySpecialFramework, but nobody you hire does. Do you really want to teach new hires your custom stack, or would you rather just hire someone who is familiar with React?But.Homegrown frameworksNobody would ever advise an aspiring race-car driver:“Oh, you just need ride along as a passenger for a few thousand rides; just watch the driver, and you’ll pick up what you need to know”Humans only improve through trial and error. Using mature software tools gives you enough experience to judge:“this library is well-designed”vs“this library an absolute flaming dumpster-fire”But it doesn’t actually give you the skills to build your own library or framework, because you have limited or no insight into the trial and error process. Building your own framework teaches you how to build a library with:Extensibility — it’s not easy to build tools which can be used in ways you didn’t foresee. Building a few tools, and seeing them used in unexpected ways, gives you the experience to judge “this class needs to be abstract” vs “no sane person would ever implement another version of this”API clarity — building an API which other people can intuitively use is hard. You won’t learn how until you do it.Documentation — likewise, it’s not easy to write documentation. You learn by doing.Backwards compatibility during upgrades — this is a super valuable skill, and incredibly hard to do. You’ll learn good practice around deprecating methods, building adapter-layers between old and new APIs, and other upgrade shenanigans.You also learn soft skills — how to take user feedback, refine it, and act on it. This is a skill you have to have to be a senior engineer or architect.SoI don’t have a magic bullet here. But I’d suggest:Don’t reinvent clear industry winners, unless you are a world-class expert in X. Don’t rebuild Hadoop, unless you are a top-10 expert in Hadoop.Don’t reinvent everything. Pick your battles.If you’re in a situation where:The framework will be a vital piece of your infrastructureThe best-in-class framework has middling or no adoption, and kinda sucksYou are willing to invest a lot of time in refining the tools, and you are willing to spend continuous effort keeping it best-in-classthen this could be an opportunity to branch out and grow your skills while building a tool specifically tailored to your current needs.But otherwise, don’t.

The Many ConnectorsThose are mostly not different implementations, just different connectors.Type B was designed for larger, desktop-sized devices. When USB came out, it was really being touted as a replacement for often huge connectors, all pretty custom for their purpose: serial ports, parallel/Centronics ports, PC mouse and keyboards, etc. The USB port delivered a better interface for pretty much everything.The Mini-B was adopted, originally, for much smaller devices, like cameras, as USB started actually becoming “universal” and consumers didn’t want to deal with legacy ports any longer. It has the exact same two signals, power, and ground that you find on Type B connectors. And while a Type B connector was only rated for 1,500 plug/unplug cycles, the Mini-B was rated for 5,000 cycles.Another thing that was added in the Mini-B generation was USB On-the-Go. USB was designed as a pure Host/Target protocol. A host device — your PC — can talk to up to 127 target devices per port (via hubs, of course). But targets can’t talk directly to targets. That was fine when your PC was talking to a mouse or keyboard, but what about a PDA or Smartphone? That’s a computer in its own right, but also has good reason to hook up to a PC via USB. So On-the-Go added one more signal, which allowed the port to decide to be either a Host or a Target. That still didn’t fundamentally change USB.Then things got smaller still, and the USB folks got together with small device makers like MP3 players and particularly smartphone makers and engineered two improvements. This yielded several things, including the Micro B connector and the “On the Go” protocol.Micro B was yet another improvement. It introduced a thinner, more rugged connector, good for 10,000 cycles. It moved the contact springs from the host/device connector to the cable, so your cable would fail, rather than your expensive device. And it was thin enough for phones.That was good until USB 3.0 (not called USB 3.1 Gen 1) came along. USB 3.0 added four new wires and an extra ground signal, and there was a Type Micro B version of this, but with that extra connector extension, it was a problem. USB 3.0 cables didn’t work with USB 2.0 ports, and the connector itself was kind of clumsy.Ultimately, we got the Type-C connector. That actually addressed a number of problems seen in the Type B connectors:Same connector for Target or Host, the protocols work that out between the two devices.Fully supports the new Power Delivery specification, making USB an even better general plug for low voltage DC power.Supports both USB 2.0, USB 3.1, and USB 3.2.No plug keying - you can plug it in either way.Still has a 10,000 cycle life.Support for “alternate modes” — Type C has a general protocol to run other signals on the same connector, allowing devices to remove other ports (DisplayPort, HDMI, etc).Smart cable tags — a Type C cable can tell your host device if there’s something special about it.In USB 1.x/2.0, power always flows from host to target. On Type C, power can run in either direction.None of this makes Type C inherently anything different than plain old USB. I have a table with a Type C connector that doesn’t support anything other than regular USB 2.0 stuff on that connector. But it’s forward-thinking, it’s going to work with all the new stuff. Type C to Type A or Type B cables or adapters allow it to work with anything else.The Several ImplementationsProtocol wise, there have been several types of USB, but they’re all backward compatible. USB 1.x had two modes: low speed, at 1.5Mb/s, and full speed, at 12Mb/s. USB 2.0 used the same signals but added the 480Mb/s high-speed mode.Then USB 3.0 came out and added four new signal wires, a pair for transmit and a pair for receive, and a 5Gb/s rate. But there’s still a separate USB 2.0 signal pair on any USB 3.0 cable or port. And next, USB 3.0 became USB 3.1, with “Gen 1” at 5Gb/s and “Gen 2” at 10Gb/s, all speeds negotiated.You can attach any USB host and target. Worse case, you may need a different cable. That’s a small price to pay for progress.