Between a windows-client and a web server, we are using bzip2 compression to squeeze off some more bytes than by just using standard gzip (the transferred data

Between a windows-client and a web server, we are using bzip2 compression to squeeze off some more bytes than by just using standard gzip (the transferred data can be huge depending on the time of the month). The client correctly sends "Accept-Encoding: x-ps-bzip2" and the server responds with "Content-Encoding: x-ps-bzip2". GData AntiVirus had a bug once in which it didn't strip out the Accept-Encoding header, tried to uncompress the servers answer (but failed) and then returned the data to the client still compressed, but with the header stripped off The client assumed uncompressed data and tried to handle the answer that way - needless to say that it had mild difficulties at parsing the compressed XML as if it were uncompressed. Of course the customer blamed us "GData is perfect - an uncle of an aunt of a friend of mine works there and it's perfect. Fix your crappy software". This was the last straw of broken AV software altering our data. We stopped using Accept-Encoding or any other additional request header (those are sometimes stripped too) and we use the User-Agent header of the client and the Server-Header of the server to signal compression. Sadly, but true, this works so far. I hate, hate, hate AV programs doing stuff like this.