Re: Cookie crumbling

FWIW, I experimented with an alternative to the existing
set-cookie/cookie mechanism that used an extension frame type with
it's own isolated compressed header block and typed codecs and
achieved a roughly 40-70% improvement (depending on cookie content) in
compression ratio over the current cookie crumbling approach.

Personally, I'd much rather -1 the cookie crumbling as a premature
optimization and explore alternative approaches later on once we have
a better defined extension model.

- James

On Mon, Oct 21, 2013 at 3:38 PM, Martin Thomson
<martin.thomson@gmail.com> wrote:
> Turns out that splitting the Cookie header into pieces improves
> compression efficiency considerably. It also turns out that we haven't
> done that, despite talking about it, testing code with it, etc...
>
> Just raised an issue to track this:
> https://github.com/http2/http2-spec/issues/292
>
> I think that we've had agreement to do this in the past, but it's been
> suspiciously absent from the compression draft.  Roberto hopes to be
> able to work this part out for the post-Vancouver/88 version of the
> drafts.
>