GSoC folks, read this: (Re: Lightweight copies/renames)

Matt Mackall mpm at selenic.com
Sat Apr 3 13:47:46 CDT 2010


On Sat, 2010-04-03 at 10:53 -0400, Paul Malmsten wrote:
> Hi Matt,

[forwarded back to the list]

> I am considering applying to the Mercurial project for Google's Summer
> of Code program, and I'm looking for more information about the posted
> projects to determine which would be a good fit for me. To this end, I
> have a few questions about the lightweight copies/renames project, and
> I learned from the mailing list that you would be the person to ask.
>       * Which skills do you think should be brought to the project in
>         order to make considerable progress?

The real challenge of GSoC is not coding, it's learning to work with a
community. And the first lesson is: communicate in public. I'm sorry
that I didn't see your first message, but I rarely answer private
correspondence about Mercurial. If I did, I'd spend my entire life
repeating myself and never get any work done.

The next big challenge is: work in public (you might begin to see a
theme here). Lots of programmers have a tendency to go off in a corner
and code until they think they've found what they think is the complete
solution. That's one of the best ways to fail at GSoC, because odds are
good you'll have overlooked some important issue and we'll send you back
to the drawing board. Alternately, you'll go off in the wrong direction
and never be able to finish. Instead, divide your project into small,
manageable parts and get feedback on each of them along the way.

And finally: be part of the community. The tendency for GSoC students
has been to work on their own projects and talk to no one but their
mentor. That's completely missing the interesting parts of the open
source experience. 

So this year, I'm hoping that students will actually engage with all
facets of the community. That means following both lists, reading
patches from other contributors, hanging out on IRC, responding to stuff
on the BTS, and engaging in things other than just your project,
including helping users. Plan on spending as much as half your time
doing these things.

Students who learn how to do this will learn a lot more valuable skills
and will be more successful than those who go off in a corner to code.
To help encourage this, this year we're going to pair students with
mentors who aren't experts in the project area so students don't have
any shortcuts around the community to getting help or getting their code
reviewed.

>       * What kinds of challenges can you foresee one encountering?

Lightweight renames is known to be a hard problem. It involves touching
code in several areas and maintaining backwards compatibility. You'll
pretty much be forced to become a Mercurial expert. And you'll need to
make sure you've got a good mental map of all the inter-related
requirements before you get too far in. We've actually had fairly
detailed discussions of what's involved here in the past and the biggest
difficulty is in managing the wire protocol changes and coordinating
with other developers who need protocol changes for their projects.

Again, a big piece of this particular problem is not the code, it's
coordinating with everyone else to understand what the code needs to do.

>       * Is this a feature frequently requested by the community?

Absolutely. Whoever successfully completes this project will be many
users' hero. And mine too - because if no one else does it, I'll
eventually be forced to.


-- 
http://selenic.com : development and support for Mercurial and Linux




More information about the Mercurial mailing list