Re: [lmi] Converting...to svn? [Was: Savannah status]

[Greg Chicares]

On 2009-06-02 12:55Z, Vadim Zeitlin wrote:
> On Tue, 02 Jun 2009 09:39:44 +0200 Vaclav Slavik <address@hidden> wrote:
> 
> VS> On Tue, 2009-06-02 at 00:26 +0000, Greg Chicares wrote:
> VS> > I can easily see several clear advantages of moving from cvs to svn.
> VS> > But, although I've spent a few hours here and there reading about
> VS> > DVCSs, I've never noticed any reason why they'd be better for lmi.
> VS> > What am I missing?
> VS> 
> VS> DVCS helps when changes propagate to upstream repository slowly or
> VS> aren't going to be accepted at all.
> 
>  I wrote that I'd be glad to reply to the question above myself but Vaclav
> did it perfectly well already and I don't have much else to say: DVCS
> indeed simplifies life a lot when you want to maintain a separate branch
> from the main/upstream one and our work on LMI has been helped a lot by
> setting up this local bzr mirror.
[...]
> VS> On the other hand, using Subversion for the central repository has an
> VS> advantage for DVCS users too: it's the common ground that DVCSs tend to
> VS> support and so you are free to use your favorite DVCS tool instead of
> VS> having to use whatever the upstream choose.
> 
>  This is a good point as well. While tailor does work for cvs->bzr
> conversion, it's a hack and has already broken down a couple of times, with
> svn things are much simpler and it's indeed "natively" supported by all 3
> main DVCS (albeit hg is slightly behind here but knowing that one of the
> main svn developers works on hgsubversion I have high hopes for the future).

Thanks for the discussion. Please tell me if you strongly disagree,
but my inclination is to move to svn now, and then analyze our
experience after that move is complete--perhaps moving to a dvcs
later. What concerns me most is not technical issues, but rather
our human limitations in mastering a different tool.

If our experience in that regard is negative, then we'll be really
glad we didn't change to a vcs that's more different from cvs than
svn is. If OTOH it goes smoothly, then we could approach a dvcs
migration with greater confidence.

Moving to svn now solves actual problems that we have today, and
it also works better with the dvcs you're using, so this step has
significant value and isn't just a throwaway experiment. And AFAIK
it's virtually guaranteed to succeed; what I don't know is how hard
it will be for us on this side of the Atlantic to adapt.