Andrew Sullivan <ajs at crankycanuck.ca> writes:
> On Wed, Jul 26, 2006 at 04:40:40PM +0200, Florian G. Pflug wrote:
>> 
>> I'd be an _optional_ feature to use. Nobody would force _anyone_
>> to use 2pc schema updates. But some people, like e.g. I, _could_
>
> But you're forcing the entire project to maintain that code.  That's
> not nothing.  It's a trade off.  
>
> Again, I'm not saying it's not worth it; I'm rather asking whether
> there isn't something that can address the actual problem you have --
> dangerous schema changes -- more completely.  I'll note that you
> simply didn't respond to my argument that perhaps this is better
> solved by improved administration tools, which is what the real
> problem is anyway.

In one sense, I'm pleased at there being a discussion of this; that
was the point to bringing it up.

I have to agree with you that the *true* problem isn't that there is
something that will be magically solved by 2PC, but rather that DDL
changes are particularly fragile, and it would be valuable to have
some improvement to that.

I brought up 2PC as a suggestion brought up, not as any sort of final
answer.

It seems to me that the discussion of this shows that 2PC isn't YET a
reasonable thing to seriously consider, since it's pretty
version-specific, forcing us to have a big hack to deal with it.

There's a potent parallel to the discussion at the Summit: lots of
people made a rumble about the notion that they'd like for DDL to be
handled better.  Unfortunately, none of that rumble included any noise
that suggested anything about how improvement might actually be
accomplished.

-> It is clear from the 2PC discussion that there are downsides to its
   use (e.g. - the potentially long period that *ALL* nodes are
   locked), and it's not clear that there are really corresponding
   merits.

-> People want "DDL to work better," but that hasn't been specified in 
   enough detail to clearly characterize what it means, let alone
   what might help solve the [amorphous perceived] problem.
-- 
select 'cbbrowne' || '@' || 'ca.afilias.info';
<http://dba2.int.libertyrms.com/>
Christopher Browne
(416) 673-4124 (land)