751 karma · joined May 2, 2013
Here was the old style:
http://website-archive.mozilla.org/www.mozilla.org/firefox_r...
And the new style:
http://website-archive.mozilla.org/www.mozilla.org/firefox_r...
The new style has since been tweaked a bit since we left Mozilla.
They also always had a list of all bugs fixed (for example, http://website-archive.mozilla.org/www.mozilla.org/firefox_r...).
Do you think we wanted to do a crazy runtime patch and didn't explore every other short/medium-term option? Do you think we wanted to write our own build system (Buck)? Or our own JS view layer (React)? Our own PHP virtual machine (HHVM)? Our own database (Presto/RocksDB)? Our own IDE (Nuclide)?
You "laughing" tells me you have never worked on a large codebase with hundreds of engineers shipping very quickly, so you can't empathize with the sort of problems that come up.
Btw, every time Google Apps hit the limit they merely bumped the limit in AOSP and dropped support for old OS versions...
1. With A/B testing and server-side feature flags they don't reflect what the user will see.
2. With technology like React Native and PhoneGap changes are decoupled from shipping an updated ipa/apk.
3. With most apps shipping quickly (some every 2 weeks) localization of release notes becomes a long pole.
4. Simple text is not effective for showing users potential benefits or training how to use something.
5. They can be misleading...saying something like "fixed a startup crash" doesn't mean for a particular user the startup crash they are experiencing is fixed.
6. More and more users have automatic update turned on and don't even see release notes is the app store UI.
7. Users don't read release notes (or anything really). HN folks may read them...normal people do not.
Do you know what version of Facebook.com you are using and all the new features they push out daily? How about Gmail? Nope, you use products for core functionality and then discover what's new in the course of actually using them and not in a text pop-up up front. This is how apps should be.
A much better and more effective strategy is contextual in-app callouts / new user experiences.
The above only applies to consumer products. I believe release notes can be valuable for enterprises/frameworks/libraries/tools/etc.
PS: I know all about the tradeoffs...I wrote release notes for Mac OS X, Firefox, and Facebook. I was the one that decided to stop doing them at Facebook for the above reasons.
http://christian.legnitto.com/blog/2016/01/02/ways-i-would-f...
https://buckbuild.com/article/exopackage.html
Although the example they show is editing a resource, so it could be fast-pathing resource-only edits (buck does that too).
More docs here:
https://sites.google.com/a/android.com/tools/tech-docs/insta...
http://blog.bolinfest.com/2015/10/hacking-on-atom-part-ii-bu...
Totally different from the "standard" way and works really well for them.
Disclaimer: I used to manage the Nuclide team at Facebook.
Also, the build tool Buck (https://buckbuild.com/) enables this because it knows the graph and associates tests with the code they test...so you even get this test minimization locally as a developer (that is, outside CI) for free. I'm sure Bazel likely has a similar feature too.
Re: pico. You'd be surprised what editors new employees and interns use when presented with a shell prompt :-D. At FB everyone is free to use their editor of choice...we don't force Nuclide on them.
Re: what's different. Xcode does Objective-C well. IntelliJ does Java well. PHPStorm does PHP well. Tons of editors do JS well. As soon as you start mixing technologies, the IDEs quickly degrade in overall usefulness and you basically have an IDE for one tech and a dumb text editor for the other. Nuclide is opinionated and provides a nice integrated end-to-end experience if you buy into the tech it supports. If you don't use any of that tech (Hack, react, buck, react native, flow, phabricator, mercurial, etc) or only use 1 language/platform at a time (like a pure objective-c iOS app) then Nuclide basically is Atom.
Nuclide's value is super apparent with React Native. You need to be able to edit JS and Objective-C and eventually Java, and it would be nice to have all the autocomplete/click-to-symbol/code search goodies for all three. Let alone debugging all of them and jumping between the stacks. No IDE provides all the features you need for a project like React Native out of the box. So we built one (leveraging Atom of course). It is magic working on a React Native app and using the same debugging UI to step through the JS, and then the Objective-C, and then using the SAME interface to debug your Hack code on the server.
Finally, over time you'll start to see a bunch of really, really smart integration with tools and libraries like Relay and GraphQL. We already have some internally--I can "click to definition" in a GraphQL query and see where it is defined in JS, and then click through and see where it is implemented in Hack on the server. Simple symbol search can't do that and spanning multiple languages and stacks is unsupported by most IDEs. Nuclide is opinionated so we can do deep integration like that...we leave the general editor-stuff to Atom proper.
Re: clumsy demo. I actually converted from vim when we started Nuclide...I even used vim for most of my JS/Android/Hack/iOS programming. For the demo I didn't install the vi bindings because I wanted fewer things that could go wrong. Because I wasn't a huge IDE user previously you see me using the mouse a ton instead of (generally) standard IDE keyboard shortcuts.
I should note that Alcatraz's clang plugin looks to have a compiled version of an old clang-format in the tree...that's likely why this doesn't work with that plugin.