I just fixed a nasty bug which caused GZIP encoded files (speak tgz etc.) to be decoded. While it'll just eat some memory on a 200 response (besides that the body is not what one would expect), it'll eat all memory on 206 (partial content) responses because the part is fed through zlib. I'll just need to revisit the HTTP RFC to check if checking for the "Vary" response header is the best bet before I drop the new release.
Also included in the post is an example script covering how one can download large files with the same, corrected extension...
