Which RDMS is best?

[ This is a repost of the following article: ]
[ From: Loki <base_at_alpha.freya.vinland.rad.mgh.mcgill.ca> ]
[ Subject: Which RDMS is best? ]
[ Newsgroups: comp.databases ]
[ Message-ID: <w70i4.3344$L41.59792_at_carnaval.risq.qc.ca> ]

Hello all,

can someone point out the best RDBMS?
currently using mSQL, which is definitely not sufficient (no transaction logging, for instance)
we have a few beta test developmental systems running Linux/SyBase 11.0.3.3

I was wondering if someone could point out the relative merits of various RDBMS's?

I need a system
that uses SQL
that can potentially be distributed
that can be replicated
that should be somewhat fault tolerant (or recoverable anyways) that will support many additions and deletions of records in tables that will support many accesses to read and search that can support many simultaneous users that can support many records
that is quick (a reason for using mSQL is that it's quick) that preferably will run under Linux
that has a PERL DBI module written for it would be nice if
it would be able to failover
have transparent access to data stored over several servers

We're currently looking into SyBase (primarily because it's available on Linux, and it is free, but we will also be looking into Oracle soon)

I've noticed that InterBase seems to support the largest databases, that Oracle and DB2 have the largest market share, that Informix and Sybase are the only other two even on the map. But PostgreSQL is also around. Ingres II seems to have many nice features (replication, shadowing built in to the product, as well as what appears to be transaction management for distributed databases) but it is still in Beta for Linux.

Anyone with an opinion?

Any other products I should look into?

Oracle
DB2
SyBase ASE (SQL Server)
Informix
InterBase
Ingres II
UniSQL
PostgreSQL

Obviously Postgres (PostQUEL based) and Ingres (QUEL based) are out of the question, because they do not support SQL. And mSQL, MySQL are disqualified because they do not have the fault tolerance we need, which goes doubly so for using DBM. ISAM based systems would not be a good idea either.

-- 

my email address is incorrect, please look at the X-De line in the header
(De - French for FROM)