Dan Fandrich wrote:
> On Wed, Dec 14, 2005 at 03:33:29PM +0100, Michael Wallner wrote:
> > BTW, shouldn't curl send "gzip,deflate" instead of "deflate,gzip"?
>
> The order doesn't matter. It might be an idea to send something like
> "gzip,deflate;q=0.5" to indicate that we prefer gzip (to avoid the
> interoperability problems of deflate), except that qvalues are not supported
> by HTTP/1.0 servers and could cause the request to fail.

RFC2616 recognises the problem of HTTP/1.0 servers, but does not
recommend against using them at all. It says:

      Note: Most HTTP/1.0 applications do not recognize or obey qvalues
      associated with content-codings. This means that qvalues will not
      work and are not permitted with x-gzip or x-compress.

I.e. it recommends not using qvalues with x-gzip and x-compress.

A server that doesn't support qvalues seems unlikely to fail the
request - I would expect it to either ignore the qvalue, or treat it
as an encoding "deflate;q=0.5" which it doesn't recognise. Neither of
those would cause the request to fail.

Do you know of any servers which actually cause such a request to fail?

-- Jamie
Received on 2005-12-15