I went to Robert Treat's session on new features in 8.4. One he liked a lot is supress_redundant_updates_trigger(), because by deploying it strategically he gets a big performance gain for very little effort. Since this was one of my contributions, I got a nice warm glow when I heard him say that. This was really not a huge effort, but it is a perfect illustration of my view that there are still many small pieces of low hanging fruit that need picking.Some of them are on the TODO list, but many are not. This one just came from some work I was doing at the time.
I did a similar type of trigger in PL/Pgsql recently, using Andrew Gierth's new version of hstore, which has a nice function to create an hstore from an arbitrary record. The reason I did this was that I needed to ignore a couple of fields when deciding whether or not to suppress the update. The new builtin trigger is very fast because it doesn't work field by field, but just does a memcmp() of the data part of the tuples.
I hope we can get a few more useful builtin trigger functions in future releases.
Since blogging the above 3 people at the conference have approached me asking for info about this feature (not because of the blog entry). They all feel it will be tremendously useful to them. Which just makes my point further.