Re: We should get 5.9** out soon

["Daniel Shahaf" <d.s@daniel.shahaf.name>]

Peter Stephenson wrote on Thu, 31 Mar 2022 11:05 +00:00:
>> On 31 March 2022 at 11:47 Daniel Shahaf <d.s@daniel.shahaf.name> wrote:
>> dana wrote on Thu, 31 Mar 2022 00:34 +00:00:
>> > On Wed 30 Mar 2022, at 11:51, Bart Schaefer wrote:
>> >> Lots of accumulated fixes, but probably the most important is to
>> >> un-break line-buffered input (workers/49792), since that problem was
>> >> introduced by the 5.8.1 security release.
>> 
>> I don't think the regression in 5.8.1 is a reason to release a 5.9; it's
>> only a reason to release a 5.8.2.  That could have a shorter pre-release
>> testing period ("soak"), and wouldn't force people to choose between the
>> regression in 5.8.1 and any as-yet-undiscovered regressions 5.9 might
>> feature (which is more likely than usual because of how large the
>> 5.8→5.9 diff is).
>
> The logic is entirely reasonable, but in practice I think stopgap fixes
> are likely just to put things off further.  I think it's worth getting
> out 5.9(?) and seeing if there's still a market for backporting non-security
> fixes to 5.8.2 --- there may not be.  (This may be the sort of thinking
> Bart had in mind.)
>

Sure, I wouldn't mind that.  I just figured it wouldn't be desirable to
delay release of the regression fix until 5.9's soak is over.  (We could
release 5.8.2 next week if we wanted to, but a 5.9 that's >2 years after
5.8 will likely want to be tested for longer than that.)

> We've never required an incompatible change to bump the major number before
> (there's often been one anyway), but given there's no single determiner it's
> something you might well take into account.

(Nothing to add to this topic at the moment.)

Cheers,

Daniel