another_anti_mysql_screed

This is part of The Pile, a partial archive of some open source mailing lists and newsgroups.



To: dbi-users@perl.org
From: Joe Brenner <doom@kzsu.stanford.edu>
Subject: [OT] mysql capabilities (was Re: RollBacks)
Date: Mon, 12 Mar 2001 17:33:10 -0800

"Tim Harsch" <harsch1@llnl.gov> wrote:

> What is the BDB table type?

Since the ACID question has been answered, I thought I'd
give you an answer on the "BDB" table type.  I'm not
going to bother being diplomatic here, so you'll have to
take it as a given that I'm a mysql-hater, and try and
compensate for my bias on your own.

For reasons that I don't entirely understand (since I
wouldn't think there's a lot of money in the game for them)
the MySQL folks have been in the business of conning web
kids into thinking that they provide real database software.
They've constantly been hassled by people who really
understand what RDMS software is about, and the first thing
that gets said is "but they don't even have transactions".

So MySQL bolted on some code from the Berkeley Database
(hence the BDB, though I've also heard this called
"Sleepycat") which does do transactions.  But this stuff
won't work with the existing MySQL tables, it only works 
with BDB type tables, and presumably with none of MySQL's 
much vaunted (and possibly illusory) speed.  So the theory
would be that using new MySQL stuff you can make a design
decisions trading off between speed and reliability.

The problems: (1) this bolted on arrangement is still a
little too new to trust it for reliability; (2) I personally
dislike having to make design decisions like this that
lock you into your original vision about what you're
doing... (in the realm of software, it makes much more sense
to me to start out using a large club to swat flies, than
it does to try and upgrade your flyswatter later when you
realize you've got bigger problems); (3) It's debateable
whether MySQL is really all that fast -- everyone who
believes that this is true seems to be basing it on the
cooked benchmarks on MySQL's site.  The industry-standard
benchmarks paid for by the postgresql people showed MySQL
performing slower even on a read-only benchmark, which is
supposed to be MySQL's forte.  

My take is that if it's at all possible, you should just
dump mysql.  If you don't have any cash, switch to
postgresql... it's not quite up to the standard of Oracle,
but the postgresql team is closing the gap really fast, and
it's a hell of lot more professional than mysql. 

===

the rest of The Pile (a partial mailing list archive)

doom@kzsu.stanford.edu