Mailing List Archive: Apache: Dev

Time for a 2.3/2.4 branch?

Index | Next | Previous | View Threaded

jim at jaguNET

Oct 4, 2009, 9:29 AM

Post #1 of 9 (588 views)
Permalink

Time for a 2.3/2.4 branch?

If we are serious about trying to get a 2.4 out this year,
what do people say about branching off trunk at this point,
so we could focus on the required checks while still allowing
trunk to continue unabated?

wrowe at rowe-clan

Oct 4, 2009, 11:21 AM

Post #2 of 9 (569 views)
Permalink

Re: Time for a 2.3/2.4 branch? [In reply to]

Jim Jagielski wrote:
> If we are serious about trying to get a 2.4 out this year,
> what do people say about branching off trunk at this point,
> so we could focus on the required checks while still allowing
> trunk to continue unabated?

-1, until we have votes for a beta/almost GA from trunk, -or- until someone
offers a breaking patch which is targeted to something later than 2.4/3.0.
That's IMHO - vetos are irrelevant to this topic. If you can point to a
recent commit as an example of what we shouldn't pick up in 2.4/3.0, you
could probably shift my opinion about this. My reasoning;

Trunk was split to allow people to make rapid progress without the overhead
of choosing the backport path and slowing down progress. In fact, progress
on httpd is mostly at a standstill by anyone other than some committed folks
happy to work through the STATUS files. The process had chased them off,
much as Aaron Bannert and others had argued. On the other hand 2.2 is very
dependable and stable as compared to other open source efforts.

So forking too early isn't healthy, and forking too late (your fear) also
isn't healthy to finally accomplish a release. Let's get to alpha and then
discuss. (Obviously, if trunk is taken in a strange direction, it's always
possible to pull the branch later from the same rev as a particular tag.)

Does this make sense?

Bill

wrowe at rowe-clan

Oct 4, 2009, 11:31 AM

Post #3 of 9 (570 views)
Permalink

Re: Time for a 2.3/2.4 branch? [In reply to]

William A. Rowe, Jr. wrote:
> Jim Jagielski wrote:
>> If we are serious about trying to get a 2.4 out this year,
>> what do people say about branching off trunk at this point,
>> so we could focus on the required checks while still allowing
>> trunk to continue unabated?
>
> -1, until we have votes for a beta/almost GA from trunk, -or- until someone
> offers a breaking patch which is targeted to something later than 2.4/3.0.
> That's IMHO - vetos are irrelevant to this topic. If you can point to a
> recent commit as an example of what we shouldn't pick up in 2.4/3.0, you
> could probably shift my opinion about this. My reasoning;

FWIW, if you question had been "if we are serious about getting 2.4 out this
year, without my suggested Apache 3.0 development getting in the way, can we
please fork now?" the answer would be different. But I'm not seeing a desire
from any folks to be working n+2 versions ahead of the next major release, and
those that might ponder it also know how to fork sandboxes :)

jim at jaguNET

Oct 4, 2009, 11:48 AM

Post #4 of 9 (573 views)
Permalink

Re: Time for a 2.3/2.4 branch? [In reply to]

On Oct 4, 2009, at 2:21 PM, William A. Rowe, Jr. wrote:

> Jim Jagielski wrote:
>> If we are serious about trying to get a 2.4 out this year,
>> what do people say about branching off trunk at this point,
>> so we could focus on the required checks while still allowing
>> trunk to continue unabated?
>
> -1, until we have votes for a beta/almost GA from trunk, -or- until
> someone
> offers a breaking patch which is targeted to something later than
> 2.4/3.0.
> That's IMHO - vetos are irrelevant to this topic. If you can point
> to a
> recent commit as an example of what we shouldn't pick up in 2.4/3.0,
> you
> could probably shift my opinion about this. My reasoning;
>
> Trunk was split to allow people to make rapid progress without the
> overhead
> of choosing the backport path and slowing down progress. In fact,
> progress
> on httpd is mostly at a standstill by anyone other than some
> committed folks
> happy to work through the STATUS files. The process had chased them
> off,
> much as Aaron Bannert and others had argued. On the other hand 2.2
> is very
> dependable and stable as compared to other open source efforts.
>
> So forking too early isn't healthy, and forking too late (your fear)
> also
> isn't healthy to finally accomplish a release. Let's get to alpha
> and then
> discuss. (Obviously, if trunk is taken in a strange direction, it's
> always
> possible to pull the branch later from the same rev as a particular
> tag.)
>
> Does this make sense?
>

Yep. My only fear, as you state, is without some clear consensus that
we want to get a 2.4 out "sometime soon", we will be stuck in that
never-ending loop of polishing the turd. ;)

paul at querna

Oct 4, 2009, 12:47 PM

Post #5 of 9 (570 views)
Permalink

Re: Time for a 2.3/2.4 branch? [In reply to]

On Sun, Oct 4, 2009 at 11:48 AM, Jim Jagielski <jim [at] jagunet> wrote:
>
> On Oct 4, 2009, at 2:21 PM, William A. Rowe, Jr. wrote:
>
>> Jim Jagielski wrote:
>>>
>>> If we are serious about trying to get a 2.4 out this year,
>>> what do people say about branching off trunk at this point,
>>> so we could focus on the required checks while still allowing
>>> trunk to continue unabated?
>>
>> -1, until we have votes for a beta/almost GA from trunk, -or- until
>> someone
>> offers a breaking patch which is targeted to something later than 2.4/3.0.
>> That's IMHO - vetos are irrelevant to this topic. If you can point to a
>> recent commit as an example of what we shouldn't pick up in 2.4/3.0, you
>> could probably shift my opinion about this. My reasoning;
>>
>> Trunk was split to allow people to make rapid progress without the
>> overhead
>> of choosing the backport path and slowing down progress. In fact,
>> progress
>> on httpd is mostly at a standstill by anyone other than some committed
>> folks
>> happy to work through the STATUS files. The process had chased them off,
>> much as Aaron Bannert and others had argued. On the other hand 2.2 is
>> very
>> dependable and stable as compared to other open source efforts.
>>
>> So forking too early isn't healthy, and forking too late (your fear) also
>> isn't healthy to finally accomplish a release. Let's get to alpha and
>> then
>> discuss. (Obviously, if trunk is taken in a strange direction, it's
>> always
>> possible to pull the branch later from the same rev as a particular tag.)
>>
>> Does this make sense?
>>
>
> Yep. My only fear, as you state, is without some clear consensus that
> we want to get a 2.4 out "sometime soon", we will be stuck in that
> never-ending loop of polishing the turd. ;)
>

start cutting alpha releases :-)

last timed we tried trunk on www.apache.org it didn't go so well...
so... we should do that again.

wrowe at rowe-clan

Oct 4, 2009, 2:28 PM

Post #6 of 9 (563 views)
Permalink

Re: Time for a 2.3/2.4 branch? [In reply to]

Paul Querna wrote:
> On Sun, Oct 4, 2009 at 11:48 AM, Jim Jagielski <jim [at] jagunet> wrote:
>>>
>> Yep. My only fear, as you state, is without some clear consensus that
>> we want to get a 2.4 out "sometime soon", we will be stuck in that
>> never-ending loop of polishing the turd. ;)
>
> start cutting alpha releases :-)
>
> last timed we tried trunk on www.apache.org it didn't go so well...
> so... we should do that again.

+1 - note that to get from alpha to GA, the biggest problem right now is the state
of docs (as Stefan hinted at). LOTS of modules are entirely undocumented. We might
want to look at these and consider dumping these from the next 2.4 release if no
documentation magically appears, courtesy of their authors.

But documentation need not block an alpha :)

My worries about going GA today mostly revolve around;

* introduction of many new hard-dependencies rather than registered functions
(one solution; for the guilty to go back and correct their designs or revert)

* problematic design of ap_internal_fast_redirect (solution; replace all calls
within httpd to ap_internal_redirect, then remove it entirely)

* problematic design of <Limit> - looks like it's time to commit <Method>
since it's probably premature to lock in all users to use mod_lua (remaining
issue; determining where the <Method> merge is evaluated)

* problematic introduction of redundant (and often error-prone) code. For example,
socache moves us partly in the right direction, but didn't remove the redundant
directive handlers from the many consumers (fortunately I'm writing an socache
consumer right now, so I'm likely to just address this)

* undocumented modules and new features (solution; document, or remove from
final release branch)

None of these are months-long efforts, it's just a matter of enough contributors
who would be looking to polish up trunk/, in proportion to those dedicating dozens
of man-months to reviewing backports to yesterday's server :)

jim at jaguNET

Oct 5, 2009, 4:06 AM

Post #7 of 9 (560 views)
Permalink

Re: Time for a 2.3/2.4 branch? [In reply to]

On Oct 4, 2009, at 3:47 PM, Paul Querna wrote:

> start cutting alpha releases :-)
>

I suggested a 2.3.3a about a month ago and the silence was deafening.

> last timed we tried trunk on www.apache.org it didn't go so well...
> so... we should do that again.
>

+1

minfrin at sharp

Oct 5, 2009, 4:34 AM

Post #8 of 9 (567 views)
Permalink

Re: Time for a 2.3/2.4 branch? [In reply to]

Jim Jagielski wrote:

>> start cutting alpha releases :-)
>>
>
> I suggested a 2.3.3a about a month ago and the silence was deafening.

As wrowe pointed out, there is a lot of work still to do - modules need
to be documented, or marked for removal if abandoned.

If we branched v2.4.x now, we would have to do this work twice, once on
trunk, and a second time on the v2.4 branch.

I don't think we're quite ready to branch trunk yet, there is still more
work to do, but cutting alphas will definitely get the momentum going.

Regards,
Graham
--

Attachments:

smime.p7s (3.22 KB)

jim at jaguNET

Oct 5, 2009, 5:08 AM

Post #9 of 9 (560 views)
Permalink

Re: Time for a 2.3/2.4 branch? [In reply to]

On Oct 5, 2009, at 7:34 AM, Graham Leggett wrote:

> Jim Jagielski wrote:
>
>>> start cutting alpha releases :-)
>>>
>>
>> I suggested a 2.3.3a about a month ago and the silence was deafening.
>
>
> I don't think we're quite ready to branch trunk yet, there is still
> more
> work to do, but cutting alphas will definitely get the momentum going.
>

I'm going to go thru and see what tasks need to be done and try
to start documenting them as release showstoppers, and in parallel
try to get a 2.3.3-alpha out.

Index | Next | Previous | View Threaded

Interested in having your list archived? Contact Gossamer Threads