On Thu, 2005-03-10 at 15:45, Jim C. Nasby wrote:
> On Thu, Mar 10, 2005 at 11:22:59AM -0600, Scott Marlowe wrote:
> > > 2. This response is alarming:
> > > Tom Lane wrote in digest V1.5092:
> > > >We are supporting Windows as a Postgres platform for the benefit of
> > > developers who want to
> > > >do testing on their laptops (and for reasons best known to themselves
> > > feel a need to run >Windows on their laptops).
> >
> > This is the second problem. Windows simply has problems that cause data
> > relibility problems that may or may not be surmountable in the future.
>
> Do you have any references to these problems? I've seen several people
> mention things like this in passing, but I have yet to see any
> specifics.
I'd have to look through the -hackers list and a few other places, but
what I remember seeing was problems in the general area of unreliable
journaling / disk syncing et. al.
It's been a while.
Plus my experience has been that Windows often behaves in unpredictable
ways when it's running under a heavy load, so I'd expect race conditions
to show up under those circumstances, and possibly corrupt data. It's
certainly been a problem for most large SQL Server installations I've
dealt with.
