Andrew Sullivan <ajs@crankycanuck.ca> writes:
> On Fri, Oct 13, 2006 at 01:35:51PM -0400, Tom Lane wrote:
>> looked reasonably robust --- ie, both safe and not full of unsupportable
>> assumptions about knowing exactly where everything actually is on the
>> disk platter. It'd still be interesting if anyone gets a new idea...
> Might it be the case that WAL is the one area where, for Postgres,
> the cost of using raw disk could conceivably be worth the benefit?
Raw disk wouldn't do much of anything to increase my comfort factor...
In practice, the answer these days for anyone who's remotely serious
is "get a battery-backed write cache", so I'm not sure how tense we
need to be about devising application-level workarounds. BBWC was
rare and expensive the last time we discussed this seriously, but
it's not so much anymore.
regards, tom lane