Re: Header Table and Static Table Indicies Switched

On 3 August 2014 11:55, Jeff Pinner <jpinner@twitter.com> wrote:

> Why would we give up 8% in compression


If that 8% is something general, then I don't think we should give it up
and I would be happy to swap back.

But you'd have to admit that your use-case of 16 headers like:

   X-Some-Very-Large-Header-Name: someSmall#

is unlikely to be a common case.   In the data I have, the usage of known
headers exceeds the usage of custom headers, so my numbers show that it is
better to favour known headers with 1B indexes.        I'd be very
surprised if general traffic was any different to that, but hopefully
somebody with access to some good generally representative data can provide
the data.... perhaps the same data set used to determine the huffman
encoding?

cheers


-- 
Greg Wilkins <gregw@intalio.com>
http://eclipse.org/jetty HTTP, SPDY, Websocket server and client that scales
http://www.webtide.com  advice and support for jetty and cometd.