Forum OpenACS Development: Re: Performance problem in PG 8.1.2
If you are doing a unique key search, then count is either 0 or 1. If you are doing a text match, then a unique key index isn't going to help much unless it is 'abc%' type. With the low size of 64chars, tsearch might be a very fast index and even pretty efficient on storage.
Be as well aware, the postgresql supports different types of indices. I have seen a couple of benchmarks in german it-newspapers comparing mysql and postgresql showing that for large data-sets postgres actually is superior.