Re: [PHP-PEAR] Re: [PHP-CVS] cvs: php4 /pear/Experimental README/pear/Experimental/HTML Menu.php

[email protected] (Jon Parise) Fri, 9 Mar 2001 17:46:45 -0500
Newsgroups php.pear,php.pear.dev
Message-ID <[email protected]>
On Fri, Mar 09, 2001 at 11:33:54PM +0100, Stig Sther Bakken wrote:

> Actually, if we start going down this route we might as well move all
> the code to the "pear" CVS module and get to the next level.

Agreed.
 
> 1. New CVS module file structure
> 
> I think the new CVS module should be structured by package names,
> rather than like a hierarchy like it is today.  The hierarchy will be
> used when files are installed.

That makes a lot of sense.  I favor this organization approach.
 
> 3. CVS tagging standard
> 
> When a release is done, the package should be tagged so it is possible
> to go back and check out the source as of a specific release later.
> The tags could be named "REL_n_n_n" (REL for release, n_n_n is the
> release/version number with all non-alphanumeric characters replaced
> by an underscore).

Yes!
 
> 4. Code staging (dealing with experimental code etc.)
> 
> We need to be able to register some additional information about each
> package.  We could use the website/database for this.  One of the
> attributes we can register is how mature the code (or a given release)
> is, and I think that could replace today's "Experimental".

Perhaps a simple file (XML?) in the top of each package directory
could contain the infortmation for that package.  It would be trivial
to periodically import the data contained therein into a database or
web page.
 
> If a package needs to fork off an experimental development cycle, I
> think CVS branches are the best thing to use, with sufficient user
> education. :-)

Branches are an excellent approach when used properly.  The
post-experimenal-cycle merges might be difficult for those who aren't
already experienced with cvs, however.

> If we decide that CVS branches are too difficult, awkward or whatever,
> it could perhaps be done on the filesystem level instead, using a
> subdirectory instead of a branch.

I prefer the branching approach over the file system approach.

As a side note, maybe Perforce would be a better choice than cvs for
this sort of repository?  Just a suggestion.  I like them both. =)

-- 
Jon Parise ([email protected])  .  Rochester Inst. of Technology
http://www.csh.rit.edu/~jon/  :  Computer Science House Member