Ok finally got my head around this :

Thanks to all those who replied !

The test case is a 10M file of 100000 records. Here are the results for
2 approaches :

i) external table followed by insert 31s
ii) embedded sql with prepared statement 23s

The external table business is quite cool - I guess it is the closest
analogy to a
bulk load tool.

With respect to my sourceforge project - I think I will amend the data
generator to be able to produce either csv or fixed style files (So I
can relatively easily include Firebird).

Finally for those of you who were interested in how Firebird compared,
here are the corrosponding Postgres results from the same box :

i) copy from file 13s
ii) embedded sql + prepared statement 90s (not a typo)


Best wishes

Mark