Re: On what we want to support: travel?

On 11/2/06, Darcy Buskermolen <darcyb(at)commandprompt(dot)com> wrote:
> > > We don't even have to win the TPC benchmarks- simply competing puts
> > > postgres in a different category.
> >
> > I agree with this point.
> >
> > And, additionally, this benchmarks could help the developers to improve
> > PostgreSQL specifically in those areas where it does not win yet.
>
> Saying that we need to join TCP to be able to improve PostgreSQL is not
> correct, we have access to a lot of "TPC like" tests
> http://pgfoundry.org/projects/tpc-w-php for example, but because it's not an
> official TPC test we can't announce it's results with any marketing weight.
> Al that TPC provides us is some official marketing buzzwords and industry

Which could very well result in expanded userbase.

A simple example: company which does not yet use PostgreSQL, and
someone suggests running some DB on PostgreSQL instead of
<some other DB>. If it involves some mission critical DB, this will be
most likely refused, because the PostgreSQL is not known to them and
untested. If it is some expendable DB, our open source competitors
will be probably at least match us performance wise, and so there is
little reason to switch to PostgreSQL. And now, having Core members
on conferences won't change that, I'm afraid. Having TPC membership
and benchmarks gives leverage (buzzwordish, but still) for convincing
sceptics into at least fully testing PostgreSQL. The "TPC-like" tests
are useless in such "political" discussions.

Then, having successfully deployed PostgreSQL in some company,
some employees will be forced to learn and to use PostgreSQL.
Some of them will like it, some even will start advocating for it.
But the first installation has to be made somehow. Sun's support
for PostgreSQL certainly helps in that area -- if you say, Hey, Sun
supports it, it's got to be worth a try, chances are people will listen
to you. TPC membership would be a similar argument, I think.

Having core members presenting in all over the Globe won't be much
help in that area. It will ceratinly help, and is needed also, but is more
of "for the future development" kind.

By the way, it could be worth to think about planning travel few months
before. For instance -- if planned early enough using, traveling using
European budget lines a SkyEurope could cost aroung $50 (usually less)
per flight. When carefully planned a person might visit few conferences
and also some "ad hoc" created meetings with local OS groups. This
would have both "bang for buck", be very tireing, and intereting (if one
likes to travel) at the same time. Having you guys flown over from the
US can be expensive (and is even more a reason for throwing you all
over the Europe ;))

> babble. Also the thought of altering PG in ways to make it perform better
> specificaly for specific benchmarks is not what we should be throwing funds
> or even more importantly developer time at.

I very much doubt any PG developer would agree on such "benchmark tuning".
Yet having PostgreSQL perform particulary badly at some benchmark is
a hint that such area needs addressing. Take a look at OSDL labs benchmarks,
they are really help identifying and eliminating bottlenecks.

Regards,
Dawid