Re: [RFC] Anti-AsyncCalls

Alex Rousskov wrote:
> On Tue, 2008-08-19 at 20:31 +1200, Amos Jeffries wrote:
>> Following some detailed discussion around the AsyncCalls issues and why
>> it blocks 3.1 release we have come to the point of withdrawing the
>> current implementation.
>>
>> With one conditional; if during the reversal TMF can complete and
>> satisfy the rest of the developers that the blocker issues have been
>> completely fixed it may be aborted.
>
> I was planning to restart this work this Wednesday. I do not think I can
> satisfy all of the developers, but I can fix what needs to be fixed.
>
> [ In the future, if "we" consider to do something drastic to the code I
> am responsible for, please consider contacting me directly before any
> decisions are made. ]
>
>> My plan for this is to;
>>
>> tag HEAD for re-reversal, then branch a duplicate of the current
>> 3-HEAD into launchpad for public preservation. This is to provide a
>> fairly stable branch where AsyncCalls can be matured past its current
>> problems before a re-merge is requested.
>>
>> Following that; the reversal of known AsyncCalls patches from HEAD.
>>
>> This to start on HEAD early September. Though initial testing will begin
>> in private by me almost immediately.
>>
>> Followed by several weeks where 3-HEAD is to be tested for stability to
>> locate any assumptions in code subsequent to AsyncCalls merge.
>>
>> When 3-HEAD is deemed stable, to make a 3.1 release with all other
>> current features. Then strike out again down the RoadMap.
>>
>> Goalpost #1 is to have 3.1.RC1 on the web before 31 October in a stable
>> and usable state.
>>
>> Maybe fanciful, but...
>>
>> Goalpost #2 is to have 3.2.PRE on the books and relatively stable with
>> limited features for performance testing by January 2009.
>
> I hope the above will not be necessary because critical bugs will be
> fixed before early September. Is September 10, 2008 an acceptable
> deadline?
>

Sept 10 is acceptable (as long as it does not slip again).
This release is really starting to pile up certain groups of users onto
HEAD code again. I'm getting deja vu on the 3.0.PRE cycle.

Amos

-- 
Please use Squid 2.7.STABLE4 or 3.0.STABLE8