Google Summer of Code & Partial cloning

[Frédéric]

On Wed, Mar 19, 2008 at 2:06 PM, Rafael Villar Burke <pachi at rvburke.com> wrote:
>  This all sounds like exciting news for mercurial users :).
>  I hope you already know about the overlay repositories concept Brendan
>  has been working on, which relates to the partial clone / shared repos
>  and obliteration features. There's some more information here:
>  http://www.selenic.com/mercurial/wiki/index.cgi/OverlayRepository

Thanks for those informations, I will test the patch and see more in
details how the concept is implemented.

On Thu, Mar 20, 2008 at 6:03 PM, Eric M. Hopper <hopper at omnifarious.org> wrote:
>  From his description I would strongly suspect narrow clones because that
>  closely mirrors the way Subversion does checking out of a subdirectory.
>  The partial history style shallow clone doesn't really have any
>  Subversion analogue at all since Subversion has a central repository and
>  therefor less need of the feature.
>
>  I too am really looking forward to this.  While having separate repos
>  for every project is generally good working practice it's not always
>  feasible, and for some huge projects, like the BSD ports tree, it
>  doesn't even really make sense.


Yes, I was toughing about both concepts but especially about narrow
partial clones, the idea was to enable developers to clone only a
subdirectory instead of the entire repos. I totally agree with the
fact that different projects must be in different repositories. But
for different reasons partial cloning is sometimes necessary. For me
it's nearly a central feature for a SCM, and I was surprised that
Mercurial doesn't support it at the moment.
To summarize, the goal of my proposal is to negate the statement "you
cannot check out only one directory of a repository" in "What
Mercurial can't do" :-)

Because of the distributed characteristic of HG, shallow partial
clones are also needed (particularly for huge projects) and to do that
the overlay concept is very interesting.

On http://www.selenic.com/mercurial/wiki/index.cgi/PartialClone I read :
"Proposal: add a new file like .hgignore to the .hg/ directory that
specifies which files should be ignored for clone/checkout."

Why not to specify only the subdir that is cloned instead of
specifying all the files to ignore ?

Cheers,
Frédéric Rechtenstein