Simon Riggs wrote:
> On Wed, 2007-08-29 at 19:51 +0200, Florian G. Pflug wrote:
>> Josh Berkus wrote:
>>> Students & Mentors,
>>>
>>> I'd like to know how all of your projects came out. I know about a few of
>>> them, but not a lot of the others:
>>>
>>> Florian -- still working on HOT Standby
>> PITR standby you mean, I think ;-)
>
> Hmmm, well if we say Hot Standby rather than HOT Standby. We need to
> differentiate what we have now ("warm") with what we hope to achieve
> ("hot"). (I do have to say I've been amazed at how many people seem to
> have confused what Florian is working on).
When I read HOT, I immediatly thought "Heap-Only-Tuples", because I've
been reading that patch, to see if we can reuse some ideas for
lazy xid assignment. So I didn't even consider that HOT could mean hot ;-)
>> I have completed an early prototype. Since the pencils-down dealine of
>> GSoC has already passed (It was 20th August), I'm currently working on
>> a patch to lazily assign XIDs. This is a generalization of a part of
>> my GSoC work, since readonly queries on PITR standby instances need to
>> be able to run without a real XID too (We cannot generate XIDs on the
>> standby, they might conflict with those generated on the master).
>>
>> If this patch should go in, it'll remove one obstacle from the way
>> to readonly-queries on PITR slaves, and make the rest of the patch
>> a great deal less intrusive.
>
> I'm very happy with what has been achieved, but we really should regard
> this as research and not committable code. In this case, the idea of
> "pencils down" doesn't make any sense.
Full Ack. The fact that what I do now won't be part of the "official SoC
outcome" anyway just made it easier for me to justify my little "lazy
XID assignment" detour.
greetings, Florian Pflug