Re: PostgreSQL vs. SQL Server, Oracle

On Wed, Oct 11, 2006 at 08:18:18PM -0400, Robert Treat wrote:
> On Wednesday 11 October 2006 12:41, David Fetter wrote:
> > On Wed, Oct 11, 2006 at 12:24:15PM +0200, Kaare Rasmussen wrote:
> > > Press coverage, an interview with Neil Matthew and Richard Stones.
> > >
> > > http://searchopensource.techtarget.com/tip/0,289483,sid39_gci1222466,00.h
> > >tml
> >
> > With friends like these...
> >
> > "In an emergency, having companies the size of Microsoft or Oracle
> > to call on may significantly mitigate that risk."
> >
> > My experiences calling these outfits in an emergency have been a
> > lot less than uniformly good, even with their top-cost levels of
> > support. Blaming one of these outfits may save some manager's
> > job, but that's not the same as actually having the emergency
> > resolved promptly, or better still, not having it happen at all.
> >
> > "First, the ability to write functions and stored procedures is
> > somewhat more limited than you would get with Oracle's PL/SQL or
> > Sybase's T-SQL."
> >
> > I don't know which languages they were looking at, but it's hard
> > to imagine how PL/SQL or T-SQL outdid PL/Perl, PL/PythonU,
> > PL/Ruby, PL/sh, etc. from a flexibility perspective.
>
> I'm not sure why people in this community are so quick to label
> anyone who is less than glowing about postgresql as "the enemy", but
> it's really annoying.

I didn't do that. I called them friends, if not very clueful ones.

> Maybe these guys were thinking about things like the ability to
> return multiple resultsets and/or the ability to do multiple
> transactions within a stored procedure;

Then they should have mentioned it. PostgreSQL has real issues, and
if they'd mentioned any one of these, it would have been reasonable.
Instead, these guys chose to spread the FUD around and call PostgreSQL
a toy.

> both of which are functionality that Oracle and SQL Server devotee's
> have been enjoying for years... (for the curious, see relevant
> threads in the -hackers archives about implementation proposals to
> add these features that as of yet have not gotten off the ground)

Part of why they haven't gotten off the ground is that it's been
possible for at least 3 years to return SETOF REFCURSOR from
functions, and there's your multiple result sets :) As far as
multiple transactions, we have had SAVEPOINTs for quite awhile and
it's possible to do 'autonomous transactions' through untrusted PLs.

I agree that none of what I just mentioned is ideal, but it means that
the capability is there, and so a lot of people who work on internals
are faced with a choice: Make existing functionality prettier, or do
things that can't currently be done. If we had a bunch more people on
salary doing this, as I hope we will soon, that choice won't be as
stark.

Cheers,
D
--
David Fetter <david(at)fetter(dot)org> http://fetter.org/
phone: +1 415 235 3778 AIM: dfetter666
Skype: davidfetter

Remember to vote!