web/links/index.htm: updated to Firefox bookmarks: PostgreSQL: query planner: documented that incorrect query plans are an ongoing bug in Postgres, because it does not support index hints and by default does not follow the join order. specifically, Postgres often does the following things in query plans which should normally never be done:

• performs a sequential scan when an index is available (because it incorrectly thinks there are too many dead rows in the index)
• uses an in-memory hash join when an index is available (because it incorrectly thinks that accessing an index on disk will be too slow)
• sorts both input tables before performing a merge join (if really no index is available, a hash join is much faster than two sorts and a merge join)
• does joins in the opposite order that they should be done (because it incorrectly thinks that the last table in the join will be filtered to fewer rows than the first table, and ignores the programmer's join order)