Re: Subversion worries

Andy Dougherty wrote:
> 
> I wonder if subversion is indeed the right tool for this job.  My
> initial reaction, based on my one-day encounter, is probably not.  Here
> are three problems I encountered with it:
> 
> 1.  There is potentially a significant installation hurdle.  Unless
>     you're on one of the few platforms for which there's a prebuilt
>     binary, you have to build it yourself.  That's ordinarily not too
>     big a deal, but when I tried this yesterday on Solaris 8, here's
>     what I encountered:
> 
>     a.  The tar file wouldn't unpack cleanly.  I get 'directory
>     checksum error'.  This has to do with filenames with length > 100
>     characters.

You're using Solaris tar ?

>     b.  The initial build went ok, but then the INSTALL file clearly
>     recommends bootstrapping from a fresh svn checkout.

I'd recommend to stick to snapshots and ignore that note in INSTALL.

>     Alas, that's
>     not a trivial job, as I was missing the following dependencies:
> 
> 	Apache Portable Runtime 0.9.4 (http://apr.apache.org/)
> 	autoconf 2.50 or newer (Unix only)
> 	libtool 1.4 or newer (Unix only)
> 	bison or yacc (Unix only)
> 	Neon library 0.23.9 (http://www.webdav.org/neon/)
> 	Berkeley DB 4.0.14
> 	Python 2.0 (http://www.python.org/)
> 	Libraries for our libraries

Yes, the svn folks aren't focused on portability. Apart from the
obligatory Linux, Windows and OS X, they don't seem to test on lots of
platforms. I tried to compile it on AIX 4.something, but I gave up at
libtool :(

(OTOH I think the Python dependency is only for their test harness)

> 2.  It's SLOW!!!!  It took me 2 hours and 24 minutes to checkout ponie.
>     One side effect of this is immediately obvious above -- it's an impediment
>     to keeping up-to-date with the rapidly changing state of parrot.

svn consumes bandwidth, yes. But more importantely the algorithms
currently used by mod_dav_svn (or whatever answers to the update
process) are not scalable when lots of files and directories are
involved. You could end up with very huge httpd processes, hence a
significant server slowdown.