I am running postgresql at version 8.1 on a windows xp desktop machine.

I then loaded 40 million rows into a table of 6 varchar fields ( The sum
total length of a row is on average about 100 characters )

If create a single btree index on one of the shortest fields ( about 6
chars long ) it takes approximately 15 minutes

On the other hand if I try to create a hash index it is still working to
create the index after 5 hours (I stopped the server at this point).
There seems to be something wrong here. I would have expected a hash
index to be quicker to create. At any rate I can't see why it should
take over 5 hours vs 15 minutes for the btree.

I read in some newgroup postings that the hash index isn't quite so
reliable, is this the problem that I have encountered, or do I need to
do anything special when creating a hash index - do I need to run on a
sever OS, for example, for this size of index.

In both cases I am using the bundled pgAdmin console to create the indexes.

I can of course use a btree index, but I wonder what went wrong when I
tried to create a hash index.