#36 fixes the problem of MD5 checksum failures when content was stored with content-encoding:gzip, but we will have a similar problem for content with content-encoding:deflate because that fix only monkey patches the decoder for the gzip case.

Two options:

1. Change accept-encoding to only be "gzip", rather than "gzip, deflate"
2. Broaden the monkey patch approach started in pull/36 to support deflate as well.

(Side note: I've read that deflate is somewhat more CPU efficient than gzip, which I think is more of a concern for browsers in general.)