Hello Tibor et al,

I know that my contribution to this discussion is absolutely marginal,
because I'm not a core contributor, and for my little patches, I'm
perfectly happy with quilt.

But, like you, I have been following all this SCM-DSCM business during
the last few years, and I have a need myself to keep track of my
scripts and utilities.  My only real experience (read+write) with SCM
was with CVS many years ago, and I have used Subversion for
downloading a couple of software packages and have some tests with
tla, Git and Mercurial.

I can subscribe your original conclussions, and I have not much to
add to what the others have said, except some random thoughts.

As we have settled with DrProject[1], a trac fork that can handle
multiple projects our wiki and task handling, and both trac and
DrProject have good integration with Subversion, in principle I was
favouring this option.

But my readonly experiences with Subversion has been quite
disappointing when I saw that, when downloading software, it cannot
keep the original timestamps of the files, something that CVS does
perfectly.  This annoys me so much!

My tests with Git and Mercurial have been driven by a quest to find a
nice tool for distributed storage for digital preservation[2].  We did
some stress-test ingesting both with 500 Gb of fat (25-75 MB) Tiff
files + some technical metadata[3].  What we found is that both are
similarly capable and both handled this load with similar timings and
overhead.  We were specially interested in Git's
content-addressable-filesystem[4] concept and the hash-tree ability to
check if two repositories are identical.  We havent' concluded
anything yet except that, compared with the mighty Git, Mercurial is
no toy either.

What else?  As there is no attractive (to me) centralised SCM option,
and being myself a low-profile developer, I'd be happy to go to a DSCM
(it can be fun), but the easier the better.  I'd be more than happy
with Mercurial.

Moreover, Mercurial has reached 1.0 this week, it has enthusiastic
followers[5], and has an Emacs frontend[6].

My cent,

Ferran


[1] http://www.drproject.org/
[2] See a summary of our preliminary findings at
    http://www.cesca.es/promocio/congressos/tsiuc2007/FerranJorba.pdf
[3] This .info file contains the md5sum and the output of
    ImageMagick's `identify -verbose'
    (http://www.imagemagick.org/script/identify.php).
[4] http://en.wikipedia.org/wiki/Content-addressable_storage
[5] For example, the first comment at http://lwn.net/Articles/274823/
[6] http://freehg.org/u/agriggio/ahg/

Reply via email to