Re: PGXLOG variable worthwhile?

On Mon, 16 Sep 2002, Peter Eisentraut wrote:

> Justin Clift writes:
> 
> > WinNT/2k/XP costs a few hundred dollars.
> >
> > MS SQL Server costs a few thousand dollars.
> 
> The places that run Windows can be categorized into three camps:  (1)
> Those that don't have a clue.  They will never run PostgreSQL.  (2) Those
> that are somehow afraid to switch to a different solution.  They will be
> even more hesitant to switch to PostgreSQL.  (3) Those that somehow like
> Windows. They will like MS SQL Server as well, no matter what we do.

I would say the only real growth market is "Those who have a clue, and are 
looking at migrating off of Windows / MSSQL to a different database."

In the case of my company, that's mostly resulted in Postgresql deployed 
on Linux and Solaris.  But I can see a use for Postgresql on Windows.  
However, for us, all our serious Windows servers have long since been 
converted to Win2K.  For all those situations, I can't imagine the 
database getting big enough and hit hard enough for pg_xlog to be a 
problem before it gets moved to a real OS.

So, by the time someone is deciding to dedicate themselves to running 
Postgresql, they've probably already decided they should run it on some 
flavor of Unix, or the slower performance of Postgresql under Windows is 
no great detriment.

Supporting a sane OS like Unix is hard enough, creating more work for the 
core developers in trying to work around a broken file system on Windows 
is not the best use of the resources available.

If and when someone running postgresql on Windows decides they REALLY need 
to move the pg_xlog somewhere else, they can either code it, or move to 
Linux.  I'd recommend moving to Linux.