Hi Markus,
>
>I checked the records in the resulting Table2.dbf. They are absolutely identical to the records our former UNIQUE index found. So, at least with these data, the MD5 hash does what I expect it to do.
>

MD5 is not the fastest one and returns 16Byte - I was planning to check out 4Byte faster returns, as they also should be faster in creating the index.
>
>BTW, I'm aware that this is not sufficient in general. But in this special case...

It might stack the odds more for winning. And if we do the first "replace" by header manipulation, using the hash directly in the index is quite might help even more.

regards

thomas