HomePhorge
Diffusion Phorge (master) Code

phorge/src/docs/roadmap.diviner
c37d90e086aemaster

Last Change
Raw File

roadmap.diviner
View Options

@title Roadmap and Status
@group intro
Insight into the direction and progress of Phabricator. This document was last
updated **June 29, 2011**.
= Status =
Phabricator is in an early release stage, but quite usable. The project is under
active development, with around 200 commits from a dozen contributors in the
last 30 days. Morale is high. A few people are tweeting about it. Someone wrote
a Wikipedia page. There are a handful of questions on Quora. As far as we know,
the product has a perfect safety record and has not caused anyone violent
physical harm.
= Current Development =
Phabricator was released about two months ago. Since then, the focus has been
on two major areas:
- **Setup/Install Process**: There was barely any install/setup documentation
when we released.
- **Pilot Installs**: Facebook was the only install when we released. We
wanted to get some early adopters to do installs and give us feedback.
The setup/install process is much better now (there's still room for work, of
course) and we've gotten a bunch of great feedback from pilot installs (and
implemented a lot of it).
We've also been focusing a lot on things that are mostly feature buildout:
- **Maniphest**: I wrote Maniphest in about 20 hours in January to track bugs
in Phabricator itself, but it seems to be getting a lot of traction. We're
doing a lot of feature buildout for it to make it more full-fledged, more
scalable, better integrated, and more useful.
- **Maturing New Applications**: Generally, there are a lot of apps
which landed recently in Phabricator's history (Maniphest, Projects, Search,
and Paste are all less than 6 months old; Herald and Diffusion are about
a year old; Differential is about 4 years old) which are missing obvious
features. We're fixing this stuff.
- **Roles and Permissions**: At Facebook, everyone basically had access to
everything but this doesn't work everywhere, especially for open source
projects. We're working on making this more granular and auditable.
- **Version 1.0**: At some point we probably need to bite the bullet, figure
out release management, and cut a release. We should probably build some
kind of update notifier before we do this. We can probably put this off for
a while longer by just declaring trunk a "kind of okay now release" since
the balance of feedback for this model has been positive (we've been able to
respond quick to a lot of bug reports and get fixes out in a few hours),
it's just off-putting to have the documentation caveating the project's
readiness so heavily.
- **Performance and Quality**: These are permanent priorities for the project,
and we're doing a lot of work to continue improving both.
= Future Development =
Most of the future work involves feature buildout. These are projects we're
thinking about in a very early stage, and may not happen or may look completely
different when we implement them:
- **Activity Streams / News Feed**: It can be difficult to get a sense of
project activity right now. Some sort of news feed is the obvious approach,
but a lot of the models for this aren't great (Google Code, GitHub,
Facebook's internal news feed) and we can probably build a more useful
product if we spend some time on it. This dataset has a pretty good mapping
to "importance" (a review is more important than a comment) which isn't as
strong in social data. Stressing that might get us somewhere useful.
- **Projects**: This is a feature which existed with a very basic
implementation at Facebook, called "tags". It had a lot of product problems
that we're experimenting with addressing by making projects more heavyweight
and structured. This feature is really bare-bones right now but seems to
be getting some traction in pilot installs and there are a lot of obvious
ways to build it out, integrate it, and make it more useful.
- **Wiki**: We're probably going to build something like a Wiki since it's
the biggest hole in the "complete package" that Phabricator presents. We
have a lot of the infrastructure we need to do this quickly and some product
ideas which could fix a lot of the problems we had with Facebook's wiki.
Biggest blocker here is coming up with a totally awesome name for it.
- **Drydock**: Build infrastructure to let Phabricator manage working copies
in a scalable way. This is a general piece of infrastructure which enables
us to build a lot of features, like: sandcastle (your changes are
automatically pushed to a machine and reviewers can access that machine to
see them), asynchronous unit testing, watir/selenium testing,
Differential-managed merging, and web bisect. This is difficult because
"scalable" is very big and it needs to shard easily across a pool of
machines. Facebook has a less general version of this which took a long time
to get working, but it solved a lot of the hard problems so it may be less
daunting for us.
- **Testing**: Phabricator has very little test coverage right now and we'd
like to improve it. But we also want to make sure we're designing the right
test environment and solving problems like database stubbing in a robust
way. Facebook ended up with some solutions in this space which had tradeoffs
and downsides we'd like to avoid.
- **Importers**: Unclear how much time we want to spend here, but providing
ways to import from other bug tracking and code review systems could lower
the barrier to adoption. But this could also be a massive timesuck.
- **Evangelism**: Phabricator had an intentionally quiet launch because the
install process wasn't any good and we wanted to get feedback (there were
other reasons, as well). It's starting to get some traction and feedback
from people have used it seems to be pretty positive. At some point it may
be appropriate to spend more time evangelizing it.
- **Mission**: Phabricator doesn't have a clear mission statement. Do we want
to develop a revenue model around it? Do we want to actively compete with
the many other products in this space? For now, improving the software is
probably the most important thing we can do to achieve any of these goals,
but we don't currently have a clear long-term vision.

Owner Packages

  • No Owners