Yes, PostgreSQL is closing in quickly on almost all the major features needed for an enterprise system. It looks like they are a version away from some of these. 7.4 will include replication/clustering, and a native Windows version, and it's already available as an alpha release (look under /dev in the FTP server).
Also, as regards the OLAP issue, there are actually some interesting OLAP-oriented tools already available in the /contrib section and at http://gborg.postgresql.org (data cube, PL/R, table cluster on index, etc...).
But, I will add a couple more missing items to Pabloj's list:
1. Nested transactions. At the moment any exception completely halts a transaction. If you are updating a million rows in a transaction by committing in small groups of 5000 (hey Mark ), this can be a real pain.
2. More capable procedural system. This is kind of hard to define easily. It's not the procedural language itself that I am talking about (although PL/PgSQL is kind of limited), but the ability for functions to interact with tables, views, and other functions is not as complete as Oracle's. This applies to all procedural languages you can use in PostgreSQL. The whole procedural system is missing such things as named parameters, passing by name, autonomous transactions, paramaterized cursors, and output parameters. These are areas I wouldn't say I am very knowledgeable on, but some Oracle experts assure me that PostgreSQL is way behind Oracle in procedural ability.
However, I am fairly sure that even these shortcomings will be dealt with quicker than anyone expects. I think the future looks good for PostgreSQL.