For anyone who may have been pondering this one (a quick post to say what was wrong would help others)... the problem was content encoding.

Webservers are supposed to look at the HTTP request headers in order to determine what kind of encoding is acceptable for the response. Normally when surfing with Firefox for example, this will be gzip/deflate (IIRC) and when requesting with the Indy components, its normally blank meaning only plain text is acceptable (IIRC).

This particular webserver was ignoring this field and was gzipping the response regardless.

As a consequence, the client needed to look at the encoding information for the response and take appropriate action... in this case, decompress the response using a gzip library such as Abbrevia.