Re: Call for Consensus: Remove "reference set" from HPACK (to address #552)

I agree to remove the reference set... but .....

I would like to see a bit more discussion of an actual draft change to do
that as I think there were a few features of the design that were driven by
the existence of the reference set.

Specifically, do we still need to copy static entries into the header set?

My understanding of that aspect feature was to control the maximum size of
the reference set for zero state implementations.  Now that there is no
reference set and the header table has it's own explicit size limit, can we
just let static references stay in the static table.

There may be some other changes needed because of this - such as putting
the static indexes below the header table indexes?

Eitherway, can we see a pull request?


On 13 July 2014 09:05, Kazu Yamamoto <> wrote:

> Hi,
> > It seems like there's broad agreement to remove the reference set from
> >
> > Jeff proposed this in <>
> to address other issues (#549-551); while it may be that it does at least
> help mitigate them, we'll consider those issues separately. We can also
> consider further modifications (e.g., requiring :-headers to be first).
> I agree to remove reference set.
> --Kazu

Greg Wilkins <> HTTP, SPDY, Websocket server and client that scales  advice and support for jetty and cometd.

Received on Monday, 14 July 2014 01:03:27 UTC