PostgreSQL 9.0 Final Release Available Now (opens in new tab)

Historically, MySQL has focused on ease of use, speed out-of-the-box, and aimed to be a lightweight, "personal" database (hence the "My"). This fast & light focus meant that certain "BigDB" features (triggers, transactions, ...) fell by the wayside.

On the other hand, PostgreSQL has historically focused on correctness and full support for advanced database features, looking to supplant Oracle and SQLServer. This meant that speed and ease of use sometimes suffered.

Today, though, both databases have mostly caught up -- MySQL has implemented most of the big ticket features that PostgreSQL had for years, and PostgreSQL (especially with the 9.0 release) has gotten far easier to use and tune, and is now as fast (or faster) than MySQL on most workloads.

The heritage still shows, at times: MySQL, even in "strict" mode, still suffers from annoying bugs and mis-interpretations of SQL features. PostgreSQL is still slow out of the box, often needing to be carefully tuned for the best performance.

Disclaimer: major PostgreSQL user and fan here. I've tried to give as objective an answer as I can.

I prefer PostgreSQL. The fact that it doesn't have many of MySQL's limitations is a big plus. Just he ability to add a column on a large table without locking the thing for a huge amount of time would be great for us. PostgreSQL can also use multiple indexes which, again, is a serious plus. The error messages that PostgreSQL returns can be much more informative than MySQL. At times when you screw up a query instead of "this is impossible" or "you can't do that", you get something more akin to "this is impossible because of X". Again, this is just my impression. Basically, PostgreSQL feels more like an open source Oracle (a grown up database) than MySQL (which doesn't feel very Oracle-y).

Now the tools are MUCH less friendly, but if you come from an Oracle (or probably DB2) world you'll be used to it. While they require more cryptic commands, they do work very well.

There is the replication thing. When we picked PostgreSQL back up this year to play with for a new little system, we discovered that they took one of the formerly external replication systems and it has been put into the tree. This is a very good step, as the replication situation was a big problem for us. Before there were little third party things (based on triggers) that you could take your chances with, and there were paid replication solutions. Having one built in is quite nice.

Of all the flamewar subjects, I think that one is just about the most tried and true one.

It's a free upgrade from warm standby, whereas SR requires a little more configuration. A good SR cluster also requires that log archiving (warm/hot standby) be set up in case the stream gets behind, otherwise you're at the mercy of whatever happens to be in the master's xlog.

Hot-Standby: Ability to query against what used to be Warm-Standby boxes which were for disaster recovery only. Streaming Replication: sending log data over sockets instead of relying on shell commands to send logfiles over.

They work together to make for happy master-slave replication.

It's worth noting this is database server level. Hot-standby copies 100% of a PostgreSQL instance. If you want to replicate individual databases on a server or individual tables in a database, Longdiste, Bucardo or Slony are still the way to go.

I use PostGIS with 9.0RC without any problems.