Jeff Mogul wrote an excellent paper that covered this problem (among
In short, do compression first, then range selection. Content-codings
are metadata about the encoding of the representation, whereas ranges
are more ephemeral.
On 2006/02/21, at 12:15 AM, Martin Thelian wrote:
If I'd like to combine Content-Encoding gzip with Content-Range.
Which offsets do I have to use in the request and response
messages. Is the offset then pointing to a position in the original
content or to a position in the encoded content? And what about the
instance-length? Is it the length of the compressed or original