Tim Hawkins wrote:
>Perhaps a better way to approach this is to specify a "plug-in" architecture
>for SVN that allows plug-in's to be called if a registered property is
>placed on a node. The plug-in could determine the behaviour of the node for
>given operations, update, commit diff etc.... It would also allow some of
>the more esoteric functions to be dropped from the core product and
>implemented as plug-ins. This would be similar to the "mod" architecture in
>apache. There would also need to be some controls to determine if the
>plug-in runs on the server and effects the repository or on the client and
>effects the wc. If specified correctly it could help to prevent cases of
>"featurisms" in the core product.
>
>
The problem is in the "specified correctly" part. Doing plug-ins the way
you describe is very, very complex from the design standpoint, and it's
definitely more than we can handle. So instead, we've approached the
problem from the other side -- by providing wrappers around the
Subversion libraries for scripting languages. We have Perl, Python and
Java based on SWIG, some (unmaintained) Ruby wrappers that could
probably be brought up to speed quickly, etc. etc. This gives you the
same kind of power.
--
Brane Čibej <brane_at_xbc.nu> http://www.xbc.nu/brane/
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Mon Jul 14 02:26:56 2003