Forum OpenACS Q&A: Can anyone here with more pull with PG please talk to him

1: Can anyone here with more pull with PG please talk to him

Posted by carl garland on 04/23/01 06:18 AM

Although it is just his opinion Phillip whom I gather was burned by
a previous Postgres installation just posted this:

PostgreSQL versus Oracle is not a sensible debate at this point. No real organization would be willing to trust its data to PostgreSQL. So if you're building applications that are going to be core to, for example, MIT's operations, you have no choice but to build them in Oracle.

This was just posted at the Ask Phillip Forum while he is allowed his
opinion it seems to carry alot of weight and I wouldn't want it to discrouage
anyone from adopting Postgres. Since I don't know him
can anyone who does please ask him to lighten up on the PG bashing.

2: Response to Can anyone here with more pull with PG please talk to him (response to 1)

Posted by Krzysztof Kowalczyk on 04/23/01 06:38 AM

The problem is that he's probably right.

He's not bashing PG as in saying "this is a bad product" but states a business reality that people who use databases for mission/business critical stuff ("real organizations" in Phil's speak) will go with an established vendor (Oracle/SQL Server) instead of trusting an open-source database, that is assuming that those organizations even know about PostgreSQL's existence. I'm inclined to believe that he's absolutely right (hell, after looking at www.postgresql.org I wouldn't trust PostgreSQL to store my appointments).

3: Response to Can anyone here with more pull with PG please talk to him (response to 1)

Posted by Stephen van Egmond on 04/23/01 07:44 AM

Who cares?

If you haven't figured it out yet, aD is only interested in Real Oragnizations because they're the only ones with the cash to pay their steep hosting and development fees. It follows that their (current) chairman is not that interested in technology their Real Organization customers don't want.

Even though he's not with aD anymore, he still needs Real Organizations to pay his bills so he can pursue his research and teaching.

Furthermore, his comment is not a senseless pgsql bash; it's entirely true. High availability is a difficult and nasty problem, which the pgsql people are beginning now (with the release of write-ahead logging) to attack.

4: Response to Can anyone here with more pull with PG please talk to him (response to 1)

Posted by Henry Minsky on 04/23/01 12:51 PM

"High Availability" is a complex goal to achieve, because it can be
nullified by any link in the whole server chain.

So even if Oracle is theoretically more bulletproof than
Postgres, if you don't have someone who is an actual expert
in configuration and recovery Oracle installations
and operating systems, then you don't actually get the
benefit of the Oracle brand name.

For example, for a long time, ArsDigita was installing Oracle
configured with archive logging turned off by default, so a "hot
backup" could not be made in case of an error that corrupted
the database files(such as filesystem
running out of space);
the only way to get a restore was using the last export dmp file,
and the dump scripts were not configured to check if this
file was being written correctly. I learned this
the hard way when a customer installation that had
been running for two years with little attention did
in fact run out of disk space, corrupted the database,
and the cron job for nightly exports had been turned off
at some point in the past. They eventually
got some procedures in place for more bulletproof recovery
procedures, when a real Oracle admin joined, but even now I don't
know if some older customers sites have been updated to new
standards.

Anyway, the point is that no one initally knew enough about Oracle
to understand how to make reliable backup architecture, even though
this was the raison d'etre for using the whole expensive
system in the first place. So it might be necessary but it is not
sufficient to say that because you have Oracle you are winning in
availability.

5: Response to Can anyone here with more pull with PG please talk to him (response to 1)

Posted by Don Baccus on 04/23/01 07:27 PM

"hell, after looking at
www.postgresql.org I wouldn't trust PostgreSQL to store my appointments"

The web site is gawdawful. It's clear that the core development team cares a lot more about hacking on the RDBMS itself than hacking on, or even supporting others who want to hack on, the web site.

It's a pity. Both Ben and I have in the past approached them about doing an OpenACS-based site but there's no interest. The folks associated with the web site itself are off by themselves in their own little lame world. They hate the notion of using AOLserver, Tcl, indeed anything that's not BSD+Apache+PHP. That would be fine if they'd put up something that's not a crock.

In case you haven't noticed, though, this site (openacs.org) is running PostgreSQL. Maybe you won't trust your appointments to PostgreSQL but by posting here you have entrusted your wisdom to PostgreSQL...

Henry ... boy, what a nightmare. I am *so* glad it was you and not me that had that joyful experience.

6: Response to Can anyone here with more pull with PG please talk to him (response to 1)

Posted by Ben Adida on 04/23/01 07:35 PM

I've posted a comment to Philip's opinion of PG. It is truly ridiculous to consider Oracle OPS a mainstream product that has any impact on the real, practical use of an Oracle database.

To be fair to the PG guys, btw, they have expressed interest in using OpenACS. They just don't have a whole lot of time to set it up. In particular, they might want to use SDM and bboards. If a few OpenACS community members want to help them out, that would be a great project to take on...

7: Response to Can anyone here with more pull with PG please talk to him (response to 1)

Posted by Pascal Scheffers on 04/23/01 07:58 PM

I would pitch in to help them setup OpenACS on their site. The biggest problem I see (with their current setup) is the site-round-robin they have. From that I get the idea that they can't even serve the static pages from one box, I'd be interested to see what happens if they run the bboards from one box (and a FAQ section from that). I guess they might be somewhat popular. If they're only concerned about download volumes, there won't be much of a problem, if the box is already running with a load avg of 20.... I wonder.

8: Response to Can anyone here with more pull with PG please talk to him (response to 1)

Posted by Mat Kovach on 04/24/01 12:47 PM

I don't know. I have pleny of customers on both Oracle (generally larger customers who already made the investment in Oracle) and Postgres (after moving them from MySQL if needed). There are risks associated with each database depending on how one sets it up. Alot of it comes down to whether the customer is willing to accept the risk or pay for the security.

I do know that Postgres is the backend for the parole telecomm app in my area, recently upgraded from Postgres 6.5 -> 7.1. Been running fine for 18 months (of course, they really don't want people to know because as they say "the stigma of using a non-commercial database". I'm constantly trying to get them to be a case study in proper DB setup and programming.

Oh well....