We're Bad At That
Architecture
No parallel query
http://archives.postgresql.org/message-id/BANLkTimwCWg6LGavaNU3YMJNcBiWppX9UQ@mail.gmail.com
http://archives.postgresql.org/message-id/1322155231200-5020742.post@n5.nabble.com
numeric is much bigger than Oracle's number
http://www.postgresql.org/message-id/50E5B990.5010101@vmware.com
Bloated tables suck
http://archives.postgresql.org/message-id/4EAC10D4.1080300@comodo.com
Backend caches require warming
Redundant updates are expensive
http://www.postgresql.org/message-id/511E469F.70308@aycan.de
AFTER trigger queues can get really big
Autovacuum doesn't know about collecting stats on inheritance trees
http://archives.postgresql.org/message-id/7806.1310233430@sss.pgh.pa.us
Executor
Indexing unordered data causes random I/O
<> is not indexable
http://archives.postgresql.org/message-id/3088.1324233086@sss.pgh.pa.us
DISTINCT + LIMIT + HashAggregate reads all input before emitting anything
Plan Types We Can't Generate
no parameterized paths aka generalized inner index-scans [fixed in 9.2]
http://archives.postgresql.org/message-id/E1RamiQ-0001jT-V9@wrigleys.postgresql.org
http://archives.postgresql.org/message-id/alpine.DEB.2.02.1204241008570.9591@krichy.tvnetwork.hu
http://archives.postgresql.org/message-id/BLU161-W1125BE7D5AED110518DEFFDD0C0@phx.gbl
http://www.postgresql.org/message-id/1454.1363046219@sss.pgh.pa.us
parameterized paths are not generated for subqueries
http://www.postgresql.org/message-id/21700.1362635555@sss.pgh.pa.us
parameterized path not used to avoid computing all windows for OVER
no merge append [fixed in 9.1]
http://archives.postgresql.org/message-id/BANLkTikntdSrhneCPzPutY9T-sAQpFXTMw@mail.gmail.com
http://archives.postgresql.org/message-id/4F54D785.6090200@clickware.de
http://www.postgresql.org/message-id/24018.1358879908@sss.pgh.pa.us
can't do batched sorts of data already ordered by leading columns