Git GSoC 2014

Git GSoC 2014

[Jeff King]

The Google Summer of Code application process is upon us. We have about
34 hours until the deadline (2014-02-14T19:00 UTC) . That's not very
much time, but I know some people have been thinking about projects for
a while, so I have hope that we can put together an ideas page.

What we need immediately is:

  - somebody to be the backup admin (I am assuming I'll be the primary
    admin, but as always, if somebody else wants to...)

  - ideas ideas ideas

    The email I sent out last week generated some response, but what I
    really need are people to write up full ideas as they would be read
    by students (and by GSoC personnel who are reading our application).

    As mentioned in a related thread, I'd really like to avoid just
    reposting proposals that didn't get any traction in previous years.

    Tomorrow I'll try to look through some of the links that have been
    posted, or collect stuff from the list. But I really need other
    people to contribute, or it's going to be a lame list.

I've put the application draft here:

  http://git.github.io/SoC-2014-Org-Application.html

You can make suggestions in this thread, or edit it by cloning:

  https://github.com/git/git.github.io

and post a patch, or send a pull request, or I should be able to give
people push access to the repo. It's a bit of an experiment, in the same
realm as the wiki we've used in the past.

The idea page is there, too, but there's nothing good on it yet:

  http://git.github.io/SoC-2014-Ideas.html

-Peff

Re: Git GSoC 2014

[Thomas Rast]

Jeff King <peff@peff.net> writes:

>   - somebody to be the backup admin (I am assuming I'll be the primary
>     admin, but as always, if somebody else wants to...)

I can be backup, if Shawn doesn't want it.

>   - ideas ideas ideas

Here's my moonshot:

--- 8< ---
Replace object loading/writing layer by libgit2

Git reads objects from storage (loose and packed) through functions in
sha1_file.c.  Most commands only require very simple, opaque read and
write access to the object storage.  As a weatherballoon, show that it
is feasible to use libgit2 git_odb_* routines for these simple callers.

Aim for passing the git test suite using git_odb_* object storage
access, except for tests that verify behavior in the face of storage
corruption, replacement objects, alternate storage locations, and
similar quirks.  Of course it is even better if you pass the test suite
without exception.

Language: C
Difficulty: hard
Possible mentors: Thomas Rast and <fill in libgit2 expert>
--- >8 ---

That absolutely requires a co-mentor from the libgit2 side to do,
however.  Perhaps you could talk someone into it? ;-)

Motivation: I believe that migrating to libgit2 is the better approach,
medium term, than rewriting everything ourselves to be nice, clean and
thread-safe.  I took a shot a while ago at making the pack reading code
thread-safe, but it's adding mess when we could simply replace it all by
the already thread-safe libgit2 calls.  It also helps shake out
incompatibilities in libgit2.

Downside: not listing "code merged" as a goal may not make the project
as shiny, neither for Git nor for the student.

-- 
Thomas Rast
tr@thomasrast.ch

Re: Git GSoC 2014

[Junio C Hamano]

Thomas Rast <tr@thomasrast.ch> writes:

> Downside: not listing "code merged" as a goal may not make the project
> as shiny, neither for Git nor for the student.

I'd actually view that as an upside. This sounds like a good first
step for a feasibility study that is really necessary.

I wonder why the handling of storage corruption and replacement
could be left broken, though. Is that because libgit2 has known
breakages in these areas, or is there some other reason?

Re: Git GSoC 2014

[Thomas Rast]

Junio C Hamano <gitster@pobox.com> writes:

> Thomas Rast <tr@thomasrast.ch> writes:
>
>> Downside: not listing "code merged" as a goal may not make the project
>> as shiny, neither for Git nor for the student.
>
> I'd actually view that as an upside. This sounds like a good first
> step for a feasibility study that is really necessary.
>
> I wonder why the handling of storage corruption and replacement
> could be left broken, though. Is that because libgit2 has known
> breakages in these areas, or is there some other reason?

It's because I don't know enough about what libgit2's state is, and I
wanted to keep the scope limited.  Naturally, the next step would then
be to implement the lacking functionality (if any) in libgit2 so that
the test suite passes.  I just don't know if that's trivial, or
something for the "if we have time" section of the project, or too much
work.

(I did do a quick "can we reasonably link against libgit2" test where I
gave git-cat-file a --libgit2 option that loads blobs with libgit2.
There are some name collisions in the git_config* identifiers that need
to be resolved, but otherwise it seems entirely possible.)

-- 
Thomas Rast
tr@thomasrast.ch

Re: Git GSoC 2014

[David Kastrup]

Thomas Rast <tr@thomasrast.ch> writes:

> Here's my moonshot:
>
> --- 8< ---
> Replace object loading/writing layer by libgit2
>
> Git reads objects from storage (loose and packed) through functions in
> sha1_file.c.  Most commands only require very simple, opaque read and
> write access to the object storage.  As a weatherballoon, show that it
> is feasible to use libgit2 git_odb_* routines for these simple callers.
>
> Aim for passing the git test suite using git_odb_* object storage
> access, except for tests that verify behavior in the face of storage
> corruption, replacement objects, alternate storage locations, and
> similar quirks.  Of course it is even better if you pass the test suite
> without exception.

[...]

> That absolutely requires a co-mentor from the libgit2 side to do,
> however.  Perhaps you could talk someone into it? ;-)
>
> Motivation: I believe that migrating to libgit2 is the better approach,
> medium term, than rewriting everything ourselves to be nice, clean and
> thread-safe.  I took a shot a while ago at making the pack reading code
> thread-safe, but it's adding mess when we could simply replace it all by
> the already thread-safe libgit2 calls.  It also helps shake out
> incompatibilities in libgit2.

That would either require forking libgit2 for Git use or stop dead any
contributions to that rather central part of the git codebase from
contributors not wanting their contributions to get reused in binary
proprietary software.

It would also mean that no serious forward-going work (like developing
new packing formats or network protocols) can be done on a pure GPLv2
codebase any more.  So anybody insisting on contributing work under the
current Git license only would be locked out from working on significant
parts of Git and could no longer propose changes in central parts.

Now this can all be repealed by the "developing the atomic bomb does not
mean that one has to use it" argument but even if one does not use it,
the world with and without it are different worlds and occupy mindshare
and suggest "solutions" and "diplomacy" involving it.

So this is definitely a large step towards a situation where erosion of
the existing license and related parts of the community becomes more
attractive.

There is the rationale "we can always say "no" at the end".  How do you
explain this "no" to the student who invested significant amounts of
work into this, in a project proposed by the Git developers?

This definitely should not be "we'll think about it if and when that
project is finished" material.

-- 
David Kastrup

Re: Git GSoC 2014

[Thomas Rast]

David Kastrup <dak@gnu.org> writes:

> Thomas Rast <tr@thomasrast.ch> writes:
>
>> Motivation: I believe that migrating to libgit2 is the better approach,
>> medium term, than rewriting everything ourselves to be nice, clean and
>> thread-safe.  I took a shot a while ago at making the pack reading code
>> thread-safe, but it's adding mess when we could simply replace it all by
>> the already thread-safe libgit2 calls.  It also helps shake out
>> incompatibilities in libgit2.
>
> That would either require forking libgit2 for Git use or stop dead any
> contributions to that rather central part of the git codebase from
> contributors not wanting their contributions to get reused in binary
> proprietary software.
>
> It would also mean that no serious forward-going work (like developing
> new packing formats or network protocols) can be done on a pure GPLv2
> codebase any more.  So anybody insisting on contributing work under the
> current Git license only would be locked out from working on significant
> parts of Git and could no longer propose changes in central parts.
>
> Now this can all be repealed by the "developing the atomic bomb does not
> mean that one has to use it" argument but even if one does not use it,
> the world with and without it are different worlds and occupy mindshare
> and suggest "solutions" and "diplomacy" involving it.
>
> So this is definitely a large step towards a situation where erosion of
> the existing license and related parts of the community becomes more
> attractive.
>
> There is the rationale "we can always say "no" at the end".  How do you
> explain this "no" to the student who invested significant amounts of
> work into this, in a project proposed by the Git developers?
>
> This definitely should not be "we'll think about it if and when that
> project is finished" material.

Yes, all of this is true.  However, you are painting a big devil on the
wall.

First, one very plausible outcome of such a project is that there is a
more narrowly defined interface to the object-reading component of the
git codebase, and that libgit2 can be "plugged in" at that interface
instead of the existing routines.  This would help both clean up our
code, and test libgit2 against existing git tests.

Your scenario above mostly applies if and when we really go the way of
my dream and scrap the code that is in git.  (I have similar long-term
dreams for other git components like ref storage and diffs, but that's
just me.)

Second, how many contributions would actually have been prevented by
GPLv2+LE licensing?

The only data I have on this is libgit2/git.git-authors, which records
who has consented for their _existing_ code to be relicensed.  I
consider this to be a higher barrier than contributing new code, since
there's no clear gain for the author in the relicensing.

That file is a sizeable list, and covers most contributions to
sha1_file.c lately:

  [git shortlog -sn --since=2.year.ago sha1_file.c, edited in the answer
  from git.git-authors]

   ok     15  Jeff King
   ok     11  Michael Haggerty
           6  Nguyễn Thái Ngọc Duy
   ok      5  Christian Couder
   ask     5  Thomas Rast
   ok      2  Brandon Casey
           2  Heiko Voigt
           2  Pete Wyckoff
           1  Felipe Contreras
           1  Joachim Schmitz
           1  Johan Herland
   ask     1  Jonathan Nieder
           1  Ramsay Allan Jones
           1  Steven Walter
           1  Vicent Marti

  (My "ask" is because of $DAYJOB legal reasons, and in fact the
  contributions above fall under the "ok before Oct 7, 2013" remark in
  git.git-authors.)

It also includes an "ok" from Nicolas Pitre, who has been the driving
force behind packv5 development.  The only thing that makes me uneasy is
that Duy is not in the list (Duy, have you been asked by libgit2 about
possible relicensing?).  Other than that, I do not see a cause for
concern.


Conversely, contributions to clean up that corner of code have not
exactly been forthcoming at a great rate in the first place.  The recent
work I can recall off hand was bug fixing and the introduction of pack
bitmaps.  The only work that I know is within reach is the packv5 drafts
that Nicolas and Duy tossed around.

(It is an odd coincidence that this thread runs in parallel with [1].
If that gains some traction, more power to them.)


So you may disagree, and other contributors should probably comment on
their stance.  But taking the above together, I think we stand to gain
more than we would lose.


[1] http://thread.gmane.org/gmane.comp.version-control.git/241965

-- 
Thomas Rast
tr@thomasrast.ch

Re: Git GSoC 2014

[Duy Nguyen]

On Sat, Feb 15, 2014 at 7:17 PM, Thomas Rast <tr@thomasrast.ch> wrote:
> It also includes an "ok" from Nicolas Pitre, who has been the driving
> force behind packv5 development.  The only thing that makes me uneasy is

Nit: pack v4. You are probably confused with index v5, which is also
cooking for a while now.

> that Duy is not in the list (Duy, have you been asked by libgit2 about
> possible relicensing?).

I don't remember. But for the record I'm OK with relicensing my
contributions in git.git for inclusion in libgit2.
-- 
Duy

Re: Git GSoC 2014

[David Kastrup]

Thomas Rast <tr@thomasrast.ch> writes:

> David Kastrup <dak@gnu.org> writes:
>
>> This definitely should not be "we'll think about it if and when that
>> project is finished" material.
>
> Yes, all of this is true.  However, you are painting a big devil on
> the wall.

[...]

> Your scenario above mostly applies if and when we really go the way of
> my dream and scrap the code that is in git.

So it's not "a big devil" I am painting but rather the consequences if
everything goes according to plan.

> (I have similar long-term dreams for other git components like ref
> storage and diffs, but that's just me.)
>
> Second, how many contributions would actually have been prevented by
> GPLv2+LE licensing?

That's not as much the question as to how many will be prevented in
future by such a step.

The libgit2 community is different from that of Git and with a different
focus.  If you take a look at its front page, you'll see statements like
"Link with open and proprietary software.  No strings attached." and
"Trusted and used in production by GitHub, Microsoft, [...]".

A fusion with this project and its aims and licensing will have
consequences regarding which developers and users are attracted to Git.

The "act first, think later" approach is not really doing anybody
favors, and I don't consider it fair to GSoC students to employ them for
making this proposal gain leverage: one should first think this through
on behalf of the project before putting students to work on this so that
they know reasonably well what will happen with their work.

The kind of "Now that we made $x do $y, we are obliged to do $z."
scenario is easy to avoid by _first_ contemplating whether or not $z is
where one wants to go.

That's not "painting a devil" but common sense.  I'm not saying that the
answer is in any way self-evident.  Merely that the best time to answer
it is _before_ getting invested.

> The only data I have on this is libgit2/git.git-authors, which records
> who has consented for their _existing_ code to be relicensed.  I
> consider this to be a higher barrier than contributing new code, since
> there's no clear gain for the author in the relicensing.

I consider it a lower barrier since the work is already done, and the
authors did not when doing it think about proprietary spinoffs.

But that's a minor point.  All I am saying is that there are different
opinions possible, and picking a particular path for future development
will in either way influence who wants to be part of the respective
communities.

-- 
David Kastrup

Re: Git GSoC 2014

[Shawn Pearce]

On Sat, Feb 15, 2014 at 4:17 AM, Thomas Rast <tr@thomasrast.ch> wrote:
> David Kastrup <dak@gnu.org> writes:
>
>> Thomas Rast <tr@thomasrast.ch> writes:
>>
>>> Motivation: I believe that migrating to libgit2 is the better approach,
>>> medium term, than rewriting everything ourselves to be nice, clean and
>>> thread-safe.  I took a shot a while ago at making the pack reading code
>>> thread-safe, but it's adding mess when we could simply replace it all by
>>> the already thread-safe libgit2 calls.  It also helps shake out
>>> incompatibilities in libgit2.
>>
>> That would either require forking libgit2 for Git use or stop dead any
>> contributions to that rather central part of the git codebase from
>> contributors not wanting their contributions to get reused in binary
>> proprietary software.
>>
>> It would also mean that no serious forward-going work (like developing
>> new packing formats or network protocols) can be done on a pure GPLv2
>> codebase any more.  So anybody insisting on contributing work under the
>> current Git license only would be locked out from working on significant
>> parts of Git and could no longer propose changes in central parts.
>>
>> Now this can all be repealed by the "developing the atomic bomb does not
>> mean that one has to use it" argument but even if one does not use it,
>> the world with and without it are different worlds and occupy mindshare
>> and suggest "solutions" and "diplomacy" involving it.
>>
>> So this is definitely a large step towards a situation where erosion of
>> the existing license and related parts of the community becomes more
>> attractive.
>>
>> There is the rationale "we can always say "no" at the end".  How do you
>> explain this "no" to the student who invested significant amounts of
>> work into this, in a project proposed by the Git developers?
>>
>> This definitely should not be "we'll think about it if and when that
>> project is finished" material.
>
> Yes, all of this is true.  However, you are painting a big devil on the
> wall.
...
> Second, how many contributions would actually have been prevented by
> GPLv2+LE licensing?

Interesting data point, I helped get libgit2 started in the first few
days of its existence and discussed the license on the mailing list. I
eventually stopped contributing, partly because of the GPLv2+LE
license it uses.

:-)

I am not as interested in using the GPL for my work as David Kastrup
is, but I wasn't really thrilled with GPLv2+LE.

Re: Git GSoC 2014

[Vicent Martí]

On Thu, Feb 13, 2014 at 10:45 PM, Thomas Rast <tr@thomasrast.ch> wrote:
> Replace object loading/writing layer by libgit2
>
> Git reads objects from storage (loose and packed) through functions in
> sha1_file.c.  Most commands only require very simple, opaque read and
> write access to the object storage.  As a weatherballoon, show that it
> is feasible to use libgit2 git_odb_* routines for these simple callers.
>
> Aim for passing the git test suite using git_odb_* object storage
> access, except for tests that verify behavior in the face of storage
> corruption, replacement objects, alternate storage locations, and
> similar quirks.  Of course it is even better if you pass the test suite
> without exception.
>
> Language: C
> Difficulty: hard
> Possible mentors: Thomas Rast and <fill in libgit2 expert>

Note that we have several people in the libgit2 team that are willing
(and excited) to mentor or co-mentor this project or any of the other
libgit2 related projects that have been proposed.

Prospective list is

- Vicent Marti
- Russell Belfer
- Ed Thomson
- Carlos Martin (past GSoC student)

So there shouldn't be any mentor shortage.

Cheers,
vmg

Re: Git GSoC 2014

[Ramkumar Ramachandra]

Jeff King wrote:
>   - ideas ideas ideas

I'll throw in a few ideas from half-finished work.

1. Speed up git-rebase--am.sh

Currently, git-rebase--am.sh is really slow because it dumps each
patch to a file using git-format-patch, and picks it up to apply
subsequently using git-am. Find a way to speed this up, without
sacrificing safety. You can use the continuation features of
cherry-pick, and dump to file only to persist state in the case of a
failure.

Language: Shell script, C
Difficulty: Most of the difficulty lies in "what to do", not so much
"how to do it". Might require modifying cherry-pick to do additional
work on failure.

2. Invent new conflict style

As an alternative to the diff3 conflict style, invent a conflict style
that shows the original unpatched segment along with the raw patch
text. The user can then apply the patch by hand.

Language: C
Difficulty: Since it was first written, very few people have touched
the xdiff portion of the code. Since the area is very core to git, the
series will have to go through a ton of iterations.

3. Rewrite git-branch to use git-for-each-ref

For higher flexibility in command-line options and output format, use
git for-each-ref to re-implement git-branch. The first task is to grow
features that are in branch but not fer into fer (like --column,
--merged, --contains). The second task is to refactor fer so that an
external program can call into it.

Language: C
Difficulty: fer was never written with the idea of being reusable; it
therefore persists a lot of global state, and even leaks memory in
some places. Refactoring it to be more modern is definitely a
challenge.

4. Implement @{publish}
(I just can't find the time to finish this)

@{publish} is a feature like @{upstream}, showing the state of the
publish-point in the case of triangular workflows. Implement this
while sharing code with git-push, and polish it until the prompt shows
publish-state.

Language: C, Shell script
Difficulty: Once you figure out how to share code with git-push, this
task should be relatively straightforward.

Re: Git GSoC 2014

[Jeff King]

On Thu, Feb 13, 2014 at 06:17:17PM -0500, Ramkumar Ramachandra wrote:

> I'll throw in a few ideas from half-finished work.

Thanks. A few comments:

> 1. Speed up git-rebase--am.sh
> 
> Currently, git-rebase--am.sh is really slow because it dumps each
> patch to a file using git-format-patch, and picks it up to apply
> subsequently using git-am. Find a way to speed this up, without
> sacrificing safety. You can use the continuation features of
> cherry-pick, and dump to file only to persist state in the case of a
> failure.

Isn't the merge backend faster? I thought that was the point of it.

> 3. Rewrite git-branch to use git-for-each-ref
> 
> For higher flexibility in command-line options and output format, use
> git for-each-ref to re-implement git-branch. The first task is to grow
> features that are in branch but not fer into fer (like --column,
> --merged, --contains). The second task is to refactor fer so that an
> external program can call into it.

I actually have this about 95% done, waiting for the patches to be
polished. So I don't think it makes a good GSoC project (it would be
stupid to start from scratch, and polishing my patches is a lame
project).

> 4. Implement @{publish}
> (I just can't find the time to finish this)
> 
> @{publish} is a feature like @{upstream}, showing the state of the
> publish-point in the case of triangular workflows. Implement this
> while sharing code with git-push, and polish it until the prompt shows
> publish-state.

I think this could be a good GSoC-sized topic. I'm going to adjust the
title to be "better support for triangular workflows". I think they may
want to examine other issues in the area, rather than drilling down on
@{publish} in particular (but ultimately, it is up to the student to
propose what they want to do).

-Peff

Re: Git GSoC 2014

[Ramkumar Ramachandra]

Jeff King wrote:
>> 1. Speed up git-rebase--am.sh
>
> Isn't the merge backend faster? I thought that was the point of it.

I suppose, but I thought git-rebase--am.sh (the default flavor) could
be improved by leveraging relatively new cherry-pick features; I
assumed that the reason it was using format-patch/ am was because it
was written before cherry-pick matured. Alternatively, can you think
of a project that involves working on the sequencer?

>> 3. Rewrite git-branch to use git-for-each-ref
>
> I actually have this about 95% done, waiting for the patches to be
> polished. So I don't think it makes a good GSoC project (it would be
> stupid to start from scratch, and polishing my patches is a lame
> project).

Oh. I look forward to using a nicer git-branch soon.

>> 4. Implement @{publish}
>
> I think this could be a good GSoC-sized topic. I'm going to adjust the
> title to be "better support for triangular workflows". I think they may
> want to examine other issues in the area, rather than drilling down on
> @{publish} in particular (but ultimately, it is up to the student to
> propose what they want to do).

That makes the project a little more open-ended then. I like it.

I was hoping you'd have more comments on "3. Invent new conflict
style". Although I'm not sure the conflict style I proposed would be
terribly useful in the general case, it'll give the student an
opportunity to look at older/ lesser-known portions of the codebase.

Re: Git GSoC 2014

[Jeff King]

On Fri, Feb 14, 2014 at 10:30:28AM -0500, Ramkumar Ramachandra wrote:

> > Isn't the merge backend faster? I thought that was the point of it.
> 
> I suppose, but I thought git-rebase--am.sh (the default flavor) could
> be improved by leveraging relatively new cherry-pick features; I
> assumed that the reason it was using format-patch/ am was because it
> was written before cherry-pick matured.

I think that's somewhat the case. But the am technique also knows a lot
of tricks that cherry-pick doesn't. For example, there is currently a
bug where "git rebase --keep-empty --whitespace=fix" silently ignores
the latter option, because the former causes it to follow a cherry-pick
code path.

So I am a little hesitant in pushing more code paths down the
cherry-pick route (though it would be OK if we correctly identified
_when_ we could use cherry-pick for performance, and only kick in then).

> Alternatively, can you think of a project that involves working on the
> sequencer?

So yeah, obviously this is all tied up with the sequencer. In the spirit
of "let's not re-propose old projects", I shied away from suggesting
"finish up the sequencer". I know that the past projects did make
progress, and that is a good thing. But I also think it doesn't make for
a good bite-sized chunk.

> >> 3. Rewrite git-branch to use git-for-each-ref
> >
> > I actually have this about 95% done, waiting for the patches to be
> > polished. So I don't think it makes a good GSoC project (it would be
> > stupid to start from scratch, and polishing my patches is a lame
> > project).
> 
> Oh. I look forward to using a nicer git-branch soon.

Actually, it is mostly about making a nicer git-for-each-ref, as I am
pulling out the ref selection code (which is more advanced in "git tag"
and "git branch") and using it everywhere. So in that sense, maybe I am
not shooting for what you want. I think you want the follow-on to that,
which is to pull out the formatting code (which is more advanced in
for-each-ref), and let it be used everywhere.

I added this into the ideas page, but noting that there were two sides
to it, and that one would need to examine and build on existing work (I
know there was some discussion and experiments on the formatting side,
too).

> I was hoping you'd have more comments on "3. Invent new conflict
> style". Although I'm not sure the conflict style I proposed would be
> terribly useful in the general case, it'll give the student an
> opportunity to look at older/ lesser-known portions of the codebase.

I almost said more. :) I am not sure I have in my mind what a useful
new format would look like, and I would worry that we are leading the
student into a bit of a trap, as they need to both code, but also invent
a new and useful format.

But one thing I was really hoping for with these project descriptions
(and I think we got) is that they are not completed project proposals.
They are the kernels of ideas that the student will need to develop into
full proposals. I would much rather have a student who reads that and
says "I have a brilliant idea for a format" and proposes it, rather than
one who blindly says "OK, I'll implement your idea". Getting the former
is much less likely, but if we do, I think it will lead to a much higher
quality project.

So I included it as-is, and I am curious to see what proposals we get.
:)

Thanks again for your list. I marked you as a potential mentor for the
conflict-style project; given the right proposal, I'd be happy to mentor
it, too (and without the right proposal, I do not think we should do it
at all). I also listed both you and me as potential mentors for
@{publish}, since we have both looked at the problem space. If you can't
make the time commitment, that's fine; I can do it (and we don't need to
decide until later anyway).

-Peff

Re: Git GSoC 2014

[Jeff King]

On Thu, Feb 13, 2014 at 04:10:37AM -0500, Jeff King wrote:

> The Google Summer of Code application process is upon us. We have about
> 34 hours until the deadline (2014-02-14T19:00 UTC) . That's not very
> much time, but I know some people have been thinking about projects for
> a while, so I have hope that we can put together an ideas page.

Just to let everybody know, the application is submitted. For reference,
I've updated the submitted application text here:

  http://git.github.io/SoC-2014-Org-Application.html

I've collected the discussion on the list on the ideas page:

  http://git.github.io/SoC-2014-Ideas.html

Google folks will be looking at it over the next week, but prospective
students probably won't see it until we are accepted to the program,
which would happen Feb 24th.

Please feel free to continue discussing or updating the ideas page in
the meantime. I think there is enough there already to show Google what
we are thinking about, but ultimately the students are the ones whom the
page is meant to serve.  Anything we can do to improve it before they
read it is a good thing.

-Peff