From what I've read, at least part of the problem is that the SQL standards are very murky and/or contradictory on the definiton of UPSERT-like functionality. And since Pg likes to be standards compliant they have focused on other areas of functionality.

Considering the rising popularity of Postgres, I'd love to see someone set up a kind of Kickstarter mechanism so the user community can fund large pieces of development work, eg partitioning , extending the work on DB checksums and so on as well as fill in any holes in the feature set like the ones you mention.

MERGE or upsert has some reasonable chance of making the next release, though of course its never guaranteed until its in there.

Right, but there are good work-arounds via stored proces and triggers. I'm using this on a couple of projects and it works perfectly.

Here is a discussion and answer in Stack overflow: http://stackoverflow.com/a/1109198/1646663