Time for 1.0

[Wilfredo Sánchez <ws...@apple.com>]

   It looks like STATUS is stale.  It doesn't show 0.9.4 was tagged.  
Can the fine person who tagged it update STATUS?

   In looking at the showstopper list, I see two API items:

     Timeout status codes aren't consistant across platforms
     apr_global_mutex_child_init and apr_proc_mutex_child_init aren't 
portable

   I posit that these can wait for the next major version which, by the 
way, doesn't have to be years away.  I also see:

     Flush out the test suite
     Eliminate the TODO's and XXX's

   These are nice-to-have's, not showstoppers, and can be worked on in 
minor versions.

   What's left?

   We have release software (HTTPd, Subversion, ?) which builds from dev 
versions of APR, and that's highly broken for many reasons.  I think 
it's high time for a version 1 release.  There is always version 2 for 
the important thing that didn't make it in version 1.

     -wsv

Re: Time for 1.0

[Justin Erenkrantz <ju...@erenkrantz.com>]

--On Friday, April 2, 2004 10:07 PM +0100 Joe Orton <jo...@manyfish.co.uk> wrote:

> which is committed to a stable API and ABI.  I don't see what's broken
> about that.  Releasing APR 1.0 today does not solve any problems:

Sure it does - we can stop being constipated about the fact that we aren't 1.0 
and start enforcing our version rules.  We *need* APR 1.0 out before we can 
even discuss doing a httpd 2.2 release - which I really would like to start 
considering soon.  It's too late for httpd 2.0 and Subversion 1.0 though. 
Let's try not to hurt ourselves anymore by being stupid and clinging to the 
fact that 1.0 must be perfect.

+1 for making whatever is living in HEAD as 1.0.  -- justin

Re: Time for 1.0

[kf...@collab.net]

Joe Orton <jo...@manyfish.co.uk> writes:
> The fact that Subversion 1.0 releases refuse to build against released
> versions of APR is a minor (and somewhat unnecessary) inconvenience
> which is trivial to work around.  Having an APR 0.9.5 release would
> solve that problem, not an APR 1.0 release.

Sorry.  Subversion 1.0 *intended* to build against a released version
of APR; we mistakenly assumed that the APR in Apache HTTPD 2.0.48 was
a release -- i.e., that the fact of its shipping with httpd meant it
was released.  Wrong assumption :-(.

Er, "The buck stops... right over there."

Re: Time for 1.0

[Joe Orton <jo...@manyfish.co.uk>]

On Fri, Apr 02, 2004 at 12:01:24PM -0800, Wilfredo Sánchez wrote:
>   We have release software (HTTPd, Subversion, ?) which builds from dev 
> versions of APR, and that's highly broken for many reasons.  I think 
> it's high time for a version 1 release.  There is always version 2 for 
> the important thing that didn't make it in version 1.

0.9.x are not "dev versions" of of APR, it's the maintenance branch
which is committed to a stable API and ABI.  I don't see what's broken
about that.  Releasing APR 1.0 today does not solve any problems:

- httpd-2.0.x releases do not build against APR HEAD

- the combination, were it made possible, of an httpd-2.0.x release and
an APR 1.0 release, would have a different module ABI to previous 2.0.x
releases, so would need a different major MMN.  This defeats the point
of the 2.0 stable branch.

The fact that Subversion 1.0 releases refuse to build against released
versions of APR is a minor (and somewhat unnecessary) inconvenience
which is trivial to work around.  Having an APR 0.9.5 release would
solve that problem, not an APR 1.0 release.

Regards,

joe