This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: Notes from the version control BOF at the summit
A few comments from the peanut gallery:
> In most new systems, the act of creating a branch is cheap, but
> parallel development is still quite expensive in terms of disk
> space. (Significantly more expensive than CVS in the case of arch
> and Subversion, I believe.)
In Subversion, with the BDB back end, each file modified on the branch
and the mainline develops two plain-text copies, which could
theoretically be expensive if you modify many files on the branch. No
one has actually measured the practical cost on a real workload,
though.
With the FSFS back end (recent development, not present in 1.0.x),
modifications on the branch and mainline are both represented as
deltas against earlier revisions (skip-deltas and delta combination
mitigate the cost), so there is no unexpected space expense associated
with branches.
> We should certainly revisit this topic in a year or so. The general
> picture will be much more clear, and there'll probably a natural
> choice by then.
The world doesn't really move so fast. Subversion will have put out a
1.1 release with some additional features (definitely the FSFS back
end and a better CVS conversion tool, probably exclusive locking,
maybe a better ACL system, but probably not merge-tracking).
Presumably tla will have made modest improvements, but the patch queue
manager, web interface, and CVS conversion facilities will remain
immature. Both will expand their user communities somewhat. darcs
may have developed a substantial user community and achieved non-toy
status; monotone will probably still be a toy. Aegis will still live
quietly off in its corner with the all features, usability, and
learning curve of a Boeing 747. CVS will continue to dominate the
day-to-day experience of most open source developers. And as today,
no one will come close to meeting all of your listed requirements.