Bruce Momjian kirjutas P, 16.05.2004 kell 22:45:
> Jan Wieck wrote:
> > We have ARC, the background writer and vacuum delay, and people even ask
> > me for backports of that (I have one for vacuum delay, but refuse to
> > make one for the others). How long do you want to delay that being ready
> > for production? Do you really think people that are suffering from the
> > fact that checkpoints, vacuum runs and pg_dumps bog down their machines
> > to the state where simple queries take several seconds care that much
> > for any Win32 port? Do you think it is a good sign for those who have
> > been our traditional Unix user base that we delay the important
> > enhancements that they need because we want to attract a lot of
> > non-professional users in Windows land? I think that is the wrong signal
> > to send. However important for marketing the Win32 port is, there are
> > other things in the pipeline that are important for those users we have
> > won already long time ago. Let's rather not lose them.
>
> Sure those are nifty enhancements, but they are not really new features.
> I would rather call them performance enhancements.
But it seems that they _are_ new enough features not to be included in
point releases (like 7.4.3).
> Sure, there are some
> folks who can't use PostgreSQL without them, but they are not like PITR
> or nested transactions where you really can't easily work around the
> limitation.
If you need a 24h non-stop database for global business, it is not
acceptable to have 30-minute periods of very slow queries resulting in
dropped connections. The only way to work around it without Jan's
patches is not to run vacuum at all, but this is a lousy longer-term
solution.
> Sure, you can work around the lack of a Win32 port with Cygwin, and
> maybe use replication in place of PITR, but the big question is are you
> hitting a large precentage of users with an enhancement.
I'm not sure that the initial version of PITR will be a good replacement
for replication.
> I am sure i
> can get some "me too's" for your improvements, but it doesn't represeent
> dramatic new functionality for PostgreSQL.
For me the vacuum delay *does* represent a "dramatic new functionality"
and I was quite disappointed that the simple version did not make it into 7.4.
> I personally don't think Win32 is enough of a new feature either, but
> others disagree.
-----------
Hannu
