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

On Fri, Jul 18, 2014 at 04:56:28PM +1000, Greg Wilkins wrote:
> Willy,
> and just some more data for you,  here is the frequencies broken up by
> field name and encoding type.

Thanks for this break-down and the other one. You convinced me :-)

> Kind of suggests that I shouldn't be
> putting content-length into the header table...
> at least not for general values:

That's a good point indeed since there's little chance it will remain
identical across multiple messages, so that would needlessly pollute
the header table after several responses. In practice, if responses
are looking similar and all headers fit into the dynamic table, it
should have no impact since each name:value pair will not be reused
and will be automatically evicted.

Best regards,

Received on Friday, 18 July 2014 07:37:08 UTC