Re: webwml in git?
* Florian Weimer <fw@deneb.enyo.de> [2008-09-17 11:30:08 CEST]:
> * Gerfried Fuchs:
> >> What happens if you exclude this data?
> >
> > I don't know exactly when this was and where. I believe it was
> > somewhere below the events pages but I can be wrong about that. Didn't
> > try to dig it up and restart the whole thing - like written, it took me
> > almost three days of (non-constant) running git-cvsimport.
>
> Okay, I will look for large files.
You will have to look for them in history, not in the checkout.
> A ChangeLog-style file is a large file which is edited in basically
> every commit. GIT handles them somewhat poorly. They tend to have a
> history overhead which is disproportionate to their actual information
> content (except when you use RCS/CVS without branching).
What's the use-case for it, in webwml? We don't have that in CVS neither?
> > To some degree yes. It should be obvious to everyone involved that a
> > conversion to a new VCS will require a new checkout and can't be done by
> > some local conversion magic, and that this new initial checkout requires
> > some bandwidth and time.
>
> It seems that the initial checkout bandwidth is around 150 MB for GIT,
> and 35 MB for CVS.
I *highly* doubt that an initial CVS checkout would take only 35 MB,
it's complete news to me that CVS is able to compress-transfer the whole
working copy anywhere near that. Please notice that the 150 MB for Git
is the whole working copy with all its history informations.
> (BTW, if we ever want a system which differentiates between production
> and beta, a changeset-based system is probably not the best choice
> because for web sites, it's common to copy select stuff from the beta
> view to the production view, instead of merging histories.)
A valid point, though it's strange to bring it up now because there
were never any approaches to go that direction in the past ...
Thanks. :)
Rhonda
P.S.: Btw., please notice that I'm subscribed to -www and don't require
the offlist Ccs, thanks.
Reply to: