W3C home > Mailing lists > Public > ietf-http-wg@w3.org > July to September 2013

Re: incrementally indexed headers should be inserted in index '0' instead of len(table)+1

From: Jeff Pinner <jpinner@twitter.com>
Date: Thu, 15 Aug 2013 15:08:08 -0700
Message-ID: <CA+pLO_iTibc+-d_NkXDAVwg9a6aT+NHbGtZxiP1foX+Gs3DH4w@mail.gmail.com>
To: Martin Thomson <martin.thomson@gmail.com>
Cc: Roberto Peon <grmocg@gmail.com>, HTTP Working Group <ietf-http-wg@w3.org>
isn't the point of delta-compression not to reference the entry again?

you reference user-agent once for example and then never again (and
hopefully it doesn't fall out of the table)


On Thu, Aug 15, 2013 at 2:57 PM, Martin Thomson <martin.thomson@gmail.com>wrote:

> On 15 August 2013 14:35, Roberto Peon <grmocg@gmail.com> wrote:
> >
> > The distance to the first appearance is algorithmically the wrong thing
> to do
>
> Actually I think that you've got the wrong end there.  You are
> thinking about the code that you built to assemble this particular
> graph.  However, if you want compression, you want to reference a
> table entry multiple times.  The longer it lives in the table, the
> better the compression.  Therefore, the longer the distance, the
> longer the period you get to reuse value over.
>
> Of course it's more expensive to scan back to the first appearance.
> But you aren't doing that when compressing, you are just scanning back
> to the point where the entry lives in the header table.  The distance
> to which is bounded by the amount of memory you are allocating to the
> table.
>
> That said, counting the number of times as opposed to the span of time
> is probably better.
>
>
Received on Thursday, 15 August 2013 22:08:35 UTC

This archive was generated by hypermail 2.4.0 : Friday, 17 January 2020 17:14:14 UTC