[tex-live] svn vs. p4

[Reinhard Kotucha]

>>>>> "Hans" == Hans Hagen <pragma at wxs.nl> writes:

  > do you have any idea why it took that much of time?

  > - too slow machine: then we need to get you a faster machine - too
  > slow (dsl) lines: then we need to get you a fast line

It seems that the bottleneck is the hard disk.  svn checks which files
have to be updated or which files conflict.

Thomas told me that you have to tell p4 in advance which files you
want to edit while svn finds changed files itself.  It scans the whole
tree therefore.  This makes it slower.

A good test is to run

  time svn status <repos>

which takes 1m32.483s (no network access required) here.  I expect
that Staszek needs at least 20 minutes.

I have a new 250GB S-ATA disk which is quite fast.  When I run svn
update twice, the second time I hear that svn is accessing the disk
again.  The second run is much faster, though.  So I suppose that
either 2GB of RAM are not sufficient to cache all the dirs or svn
writes some meta data even if nothing has to be updated.

Staszek's DSL connection is certainly not slower than mine, maybe even
a bit faster.  At least it seems that I need more time to connect to
the server.

Hans, I think that if we will use svn in the future, it's best if
Staszek could get a reasonably fast machine.  There are a few other
things to be considered: The .svn files in the ls-R databases might
slow down things as well and the Build script will benefit from a
faster machine too.  And what Staszek did in the past will justify
some support anyway.

As I said before, I can live with p4 as well.  I didn't use it before
but I installed a client already.  And I even can't tell now how much
I can contribute to TeXLive this year.  Don't care about me.  So if we
find a solution which makes Staszek happy, that would be a big step.

Though Staszek and Karl did most of the work, I think that some other
people contributed and it would be nice to get some feedback from
them.  Gerben mentioned the problem to create binaries on old
machines, which is a very important point.  It also affects Staszek
when he builds binaries using old versions of libc.

Would it be helpful to have two independent repositories, one for
system independend stuff (texmf-dist and texmf-doc) and one for system
dependent stuff (texmf and source)?  In a previous mail I mentioned a
workaround, but a workaround is not always a solution, at least not a
good one.

It seems that svn is significantly slower than p4 though many people
prefer svn.  It obviously requires more disk space.  It would be a
pity if we decide to use svn now and it turns out later that we cannot
have binaries for old systems because of hardware limitations.

We need more feedback from all the people involved in TeXLive
production somehow.  The question is: Can you live with svn or not.

I'm a bit concerned that a decision is made now and people run into
problems later.

Regards,
  Reinhard

-- 
----------------------------------------------------------------------------
Reinhard Kotucha			              Phone: +49-511-4592165
Marschnerstr. 25
D-30167 Hannover	                      mailto:reinhard.kotucha at web.de
----------------------------------------------------------------------------
Microsoft isn't the answer. Microsoft is the question, and the answer is NO.
----------------------------------------------------------------------------