"Daniel L. Rall" wrote:
> Bill Tutt wrote:
> > Greg Ward wrote a Python script to munge CVS's metadata to auto-repoint the
> > checked out tree to Python's new CVS tree at sourceforge after it moved away
> > from CNRI.
> > So yes, people do manipulate the metadata files.
> I do this constantly, often on the command line. Here's an example of
> replacing the CVSROOT for a project:
> for d in `find . -type d -name CVS`; do echo
> ':pserver:firstname.lastname@example.org:/cvs' > $d/Root; done; unset d
Such operations should be part of the working copy library API.
I know CVS doesn't provide a command to change the repository root,
but there's no reason why Subversion shouldn't do so.
In my (not so humble) opinion, metadata should _only_ be manipulated
via the Subversion client library. Tools that operate on the working
copy can issue SVN API calls to read the metadata instead of parsing
it themselves. This has several benefits:
- We can change the format and contents of the metadata
without breaking every tool. We only have to maintain
- We can change the _location_ of the metadata. We can even
allow different kinds of checkouts, e.g., one kind could
store metadata locally in the working copy, another could
leave it on the server.
(Why? For instanc, someday we may want to implement something
similar to ClearCase's dynamic views. Keeping metadata on the
server would make that easier.)
Branko Čibej <email@example.com>
HERMES SoftLab, Litijska 51, 1000 Ljubljana, Slovenia
voice: (+386 1) 586 53 49 fax: (+386 1) 586 52 70
Received on Sat Oct 21 14:36:07 2006