A group of folks is going to get together for some Loop hacking tomorrow
(Saturday) in London and are going to need to dig into the front-end
pieces of MLP.
In the interest of:
* getting people moving fast
* placing what we think are good bets
* validating those bets during MLP
* failing them fast if they turn out to be problematic
I've collected a list of decisions, based on a lots of conversation with
a variety of people, and I've reviewed that list with Adam, and I'm
attaching it below. This week got complicated, which made it
impractical to send these out one-by-one with more time before the hack
session this weekend, for which I apologize.
The bugs have varying amounts of context in them, (some of it is still
in notes or people's heads); I'll be spending the rest of today going
through them, adding more context, and noting where context is missing.
Please keep in mind that we're trying to jump-start things, and these
decisions are specifically for starting MLP: a lot of the point of MLP
is validate these sorts of decisions and failing them fast if they
don't work out. If you have a concern that is so strong that you're sure
we should reconsider despite that, please find and talk to me or Adam
before digging into the bugs for general discussion.
Localization (bug 972884):
* webl10n, same as pdf.js in Firefox, on the web, and in b2g. Also used
for b2g itself.
* leverages our existing localization tools and community
Stand-alone and in-browser unit-testing framework (bug 976789)
* mocha & chai, already used by b2g for gaia as well as the loop server
* good at structuring large suites in maintainable ways
* fast and very efficient toolchains in and around these tools
* after talking to folks on the a-team, it appears that it should be
straightforward to integrate with the Tbpl infrastructure
CSS toolkit for standalone and in-browser code: (bug 976854 and bug 976857)
* None; easy to mix-in later if we want something later
MVC infrastructure (bug 975548)
* Well-understood/tested code structuring for moving fast
* Many B2G apps have rolled their own, and it hasn't been easy.
* Backbone + dependencies optimized for mobile (zepto/underscore) CSS
toolkit for standalone and in-browser code
Call Notification (bug 976789)
* SimplePush: as far as we can see, this is the only easily-scalable
solution that exists
Client-driven end-to-end testing framework for both browser and
standalone (bug 976134, bug 976114, bug 972026)
* Marionette generally
* mix-in Steeplechase for tests that need NAT traversal
** will need work from Ted, which he sounds open to doing
Dropdown panels and chat windows (bug 971987, bug 976358)
* SocialAPI (with whitelist)
* This seems like the fastest way to stand up the user-experience that
we're looking for.
Thanks,
Dan
_______________________________________________
dev-media mailing list
[email protected]
https://lists.mozilla.org/listinfo/dev-media
