RTC on 0.9.x? was Re: svn commit: r161087 - in apr/apr-util/branches/0.9.x: CHANGES include/apr_reslist.h misc/apr_reslist.c

[Paul Querna <ch...@force-elite.com>]

Paul Querna wrote:
> William A. Rowe, Jr. wrote:
> 
>> At 01:38 PM 4/12/2005, pquerna@apache.org wrote:
>>
>>> Author: pquerna
>>> Date: Tue Apr 12 11:38:21 2005
>>> New Revision: 161087
>>>
>>> URL: http://svn.apache.org/viewcvs?view=rev&rev=161087
>>> Log:
>>>
>>> Backport apr_reslist_timeout_set and apr_reslist_invalidate.
>>>
>>> These functions have been in 1.x for a long time.  I am personally 
>>> using them in apr_memcache, and lacking apr_reslist_invalidate was 
>>> the only reason I had to require APR 1.x.
>>
>>
>>
>> Please do NOT change the API on released branches without discussion.
>> They are RTC.
> 
> 
> I believe that RTC is present on HTTPD 2.0.x.
> 
> I believe APR 0.9.x is under CTR.

Looking back at recent commits, 'apr_threadattr_stacksize_set' was added 
  to 0.9.x in January. This was a back port from 1.x.  I find no record 
of any voting.

This would of fit your criteria of an API change/addition on a released 
branch.

I am open to putting 0.9.x under RTC officially, but it does not seem 
this has been an enforced policy. (Hence, it has always seemed CTR to me).

The STATUS file in APR and APR-Util does not say 0.9.x is under RTC. The 
httpd STATUS file for 2.0.x does mention that changes must be voted on.

If APR and APR-Util in the 0.9.x branch should be under RTC, that is 
fine with me.  I would prefer an official stance, and for that to be 
reflected in the STATUS files, rather than my commit getting picked out, 
when others have done CTR before.

-Paul

-Paul

Re: RTC on 0.9.x? was Re: svn commit: r161087 - in apr/apr-util/branches/0.9.x: CHANGES include/apr_reslist.h misc/apr_reslist.c

["William A. Rowe, Jr." <wr...@rowe-clan.net>]

At 05:34 PM 4/13/2005, Justin Erenkrantz wrote:
>--On Tuesday, April 12, 2005 4:28 PM -0400 Cliff Woolley <jw...@virginia.edu> wrote:
>
>>I was also under the impression that all branches of APR were CTR.
>>
>>However, I agree that discussion on API changes would be good, even in a
>>CTR system.  That has always been the basic idea with CTR -- you can
>>commit without prior review, but it is understood that anything major will
>>be discussed ahead of time.
>
>+1.  -- justin

Well put, Cliff.  Thank you.

Bill

Re: RTC on 0.9.x? was Re: svn commit: r161087 - in apr/apr-util/branches/0.9.x: CHANGES include/apr_reslist.h misc/apr_reslist.c

["William A. Rowe, Jr." <wr...@rowe-clan.net>]

Justin Erenkrantz wrote:
> --On Tuesday, April 12, 2005 4:28 PM -0400 Cliff Woolley 
> <jw...@virginia.edu> wrote:
> 
>> I was also under the impression that all branches of APR were CTR.
>>
>> However, I agree that discussion on API changes would be good, even in a
>> CTR system.  That has always been the basic idea with CTR -- you can
>> commit without prior review, but it is understood that anything major 
>> will
>> be discussed ahead of time.
> 
> +1.  -- justin

I don't agree with a 'hard and fast' rule, that's broken httpd horridly.

But note we were going to bump 'version minor' every time we introduced
new functions - at least that's what Greg's original versioning doc had
suggested.  "Changes" require a version major bump (e.g. it doesn't
match a previous, released API, causing ABI breakage.)

AFA APR[-FOO] 0.9, all bets are off, since we had no rules back then.

So, API changes are always good, they either bump minor (new feature)
or bump major (which should be managed by the whole project so we can
get in and deprecate dead functions).

But we gotta work as a team - and more rules don't help, cooperation
does.  This is why I brought up the apr_dso and apr_app functions to
determine full paths to an arbitrary loaded library/application's path.
Better to warn all, get as much feedback as possible before doing 'new
stuff'.

Bill

Re: RTC on 0.9.x? was Re: svn commit: r161087 - in apr/apr-util/branches/0.9.x: CHANGES include/apr_reslist.h misc/apr_reslist.c

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

--On Tuesday, April 12, 2005 4:28 PM -0400 Cliff Woolley 
<jw...@virginia.edu> wrote:

> I was also under the impression that all branches of APR were CTR.
>
> However, I agree that discussion on API changes would be good, even in a
> CTR system.  That has always been the basic idea with CTR -- you can
> commit without prior review, but it is understood that anything major will
> be discussed ahead of time.

+1.  -- justin

Re: RTC on 0.9.x? was Re: svn commit: r161087 - in apr/apr-util/branches/0.9.x: CHANGES include/apr_reslist.h misc/apr_reslist.c

[Cliff Woolley <jw...@virginia.edu>]

On Tue, 12 Apr 2005, Paul Querna wrote:

> > I believe APR 0.9.x is under CTR.

I was also under the impression that all branches of APR were CTR.

However, I agree that discussion on API changes would be good, even in a
CTR system.  That has always been the basic idea with CTR -- you can
commit without prior review, but it is understood that anything major will
be discussed ahead of time.

--Cliff