Amazing work guys! Thanks to the shiny new API in l20n.js, the Music NGA re-write was able to take full advantage of all the goodness it brings, including RTL-compliant time formatting and the client/service split for views via bridge.js.
Also, many, many thanks for all the perf improvements! :-) -Justin > On Nov 4, 2015, at 12:30 PM, [email protected] wrote: > > Hi Gaiaers! > > I'd like to start a new tradition of providing you with a summary of work > that has been done during the a given cycle in my module[*]. > > It may be surprising for people less familiar with complexity of the l10n > ecosystem to notice that our L10n team has been so visible in the number of > commits - both me and :stas accruing over 120 commits in this cycle placing > us both in top 20 commiters. > > The reason is: L10n is complex. It's also severely under-developed compared > to most layers of modern app systems. And that is both, a curse and a > blessing. A curse because we have no-one to learn from. Nobody has figured it > out, no GUI toolkit has a good l10n system and we're with l10n more or less > where the webdev world was with HTML4 and PHP. > > The blessing is that we have an opportunity at Mozilla to push the state of > the whole industry. It's a niche where with limited resources that we have, > we can make the Web Stack better than any other stack. > I humbly believe that we are doing it, and 2.5 is an important milestone on > that route. > > > The biggest accomplishment of this cycle for us is the introduction and > deployment of new L10n API. This API is based on years of research work by an > L10n team in Mozilla led by Axel Hecht. When we took over L10n module in > Firefox OS, we started moving the module and refactoring the codebase to work > in the paradigms of that approach. > > And at the beginning of 2.5 cycle Stas landed revision 3 of our API in Gaia > source. Right now it's opt-in: the code needs to link to > shared/js/intl/l20n.js to use it. > > For the rest of the cycle we improved the new codebase vastly improving > performance and memory consumption. At this point L20n 3.0 beats the previous > API in features, security, performance and memory. > > This, and a lot of refactoring of the code to be more asynchronous, allowed > us to move the first four Gaia applications to it - Music, FM, SMS and FTU > are fully transitioned and running on top of the new API. > > Other highlights: > > * Significant improvements in our test coverage, asynchronous non-racy APIs > and security of HTML l10n > * Huge progress in reduction of our technical debt. We removed thousands of > lines of complex JS code replacing them with simple declarative HTML > attributes > * DOM Overlays API allows us to localize Document Fragments > * Bidi markers make composed localized strings work well in LTR/RTL scenarios > * We transitioned whole Gaia to use Intl API for date and time formatting > which significantly improves Firefox OS user experience in non-latin based > languages > * We moved to use the same Intl API in our composed localization strings for > number formatting, so any numbers passed to l10n as variables will be > formatted properly for the current locale > * After trying multiple solutions we settled on an amazing module bundler > internally, rollup.js, which helped us save memory. > * We leveraged the new internal architecture to properly use client-service > bridge architecture in multi-view apps (see Music) > * Made our build system much more strict about l10n errors allowing us to > catch many potential race conditions with localization resources > * Introduced first experimental version of our future localization format - > l20n > * We deployed mozIntl API that is a staging ground for future JavaScript Intl > API > * Lots of work on ECMA TC39 group to get a lot of features needed for Firefox > OS into Intl API rev. 3 (ES7). > * migrated our pseudo-locales to work well with Intl API (following Google > pseudo-locales rather than Microsoft pseudo-locales) > * We're helping with a lot of performance related work, with both me and > :stas becoming peers of the Performance Module and improving our ability to > make Gaia apps start fast in any locale > * We are working with JS/Gecko and Gonk teams to fix long standing > time/timezone/date problems to make our platform work well across timezones. > * deprecation of the old l10n_date.js library in favor of Intl API and mozIntl > * removal of over 400 calls to the deprecated mozL10n.get method! > * some ground work to enable our whole build/l10n workflow to handle our new > l10n format > > > That's a lot of work. Over 120 commits in gaia 2.5 is just the visible part > of the iceberg with over 300 commits to our l20n.js repository and lots of > work going to various l10n/intl polyfills, ECMA specs and other repos. > > With 2.5 being done, we already started working on 2.6 cycle which we're very > excited about because it feels like we have less and less of cleanup work, > and we start to work on real improvements and new features! We still have > some debt, but we believe 2.6 may be *the* cycle in which we'll finish that > part of our work. > So, what's in our plans? Here are some teasers: > > * Final rush to remove all mozL10n.get calls (almost 500 calls to remove). > Most of that work is in System and Communications > * Integration tests - we now have a framework for that and we can start > testing our HTML/DOM API in the browser > * Move (almost) all apps to l20n.js and deprecate l10n.js > * Introduce declarative Intl API for HTML > * Try to remove build time optimizations for l10n > * More work on advancing Intl API and getting 3rd edition of ECMA 402 to have > everything we learned we need to make Firefox OS look good > * Start using our new localization format and introduce new l10n features > that it allows for > * Move language packs to our build system > * We are starting to look at Web Extensions and browser.html and poke them > with a probing stick > * We are experimenting with exposing l20n.js 3.x to the Web community. > Supporting some older browsers like IE11 is the current crux. > > I don't know how many of those goals we will be able to accomplish in this > cycle and how many will have to catch the next train, but with those > features, I believe that we will have the best multi locale UI platform in > the World. And that's a goal worth pursuing! > > Thank you to all reviewers, committers, allies, RTL team, and the amazing > sheriffs who had to back out a few eagerly landed patches. We're working on > it! :) > zb. > > > [*] I believe it would be valuable for us to start providing such per-cycle > summaries rather than per-quarter summaries. Quarters are artifacts of our > management structure, not our project structure. > _______________________________________________ > dev-fxos mailing list > [email protected] > https://lists.mozilla.org/listinfo/dev-fxos _______________________________________________ dev-fxos mailing list [email protected] https://lists.mozilla.org/listinfo/dev-fxos
