> That's right.. but this is indeed a bad solution. A 
> slightly better one would be to ensure that 404 pages are 
> at least 512 bytes-long. Or even to respond with a 3xx code 
> (redirect) and a standard, UNIQUE, 404 page.
> 
> But to give 200 errorcodes in such cases is indeed a bad 
> idea (especially for the loop problem)

I had a nightmare of a site recently where the "404" page returned a 200 code. 
The problem was, though, that their "404" page was a full page around 17KB
with many links to other files which were ALSO 404 (so all those of course had
200 codes too).  Basically the mirror was in a crazy loop, and I had to add
many many filters so that the HTTrack never reached a "404" :(