W3C home > Mailing lists > Public > ietf-http-wg@w3.org > January to March 2014

Re: HPACK: replace Huffman by possibly faster and more efficient FSE (Finite State Entropy)

From: Michael Sweet <msweet@apple.com>
Date: Mon, 06 Jan 2014 11:31:23 -0500
Cc: ietf-http-wg@w3.org, Roberto Peon <fenix@google.com>, Hervé Ruellan <herve.ruellan@crf.canon.fr>
Message-id: <A8802139-AC3B-4F95-9F20-2007CA50EFD8@apple.com>
To: Frédéric Kayser <f.kayser@free.fr>

Certainly looks interesting, although IMHO all of this focus on compression for HTTP/2.0 is premature - yes, we can reduce the size of the request/response headers even more, but once you get past optimizing the initial request/response (initial page load time) any further gains are just "nice to have".

Better to have a simpler HTTP/2.0 that achieves the chartered goals than to create a much harder to implement HTTP/2.0 that offers only a modest improvement in perceived performance/responsiveness.  And IMHO adding Huffman or FSE falls into the "harder to implement for a modest improvement" category.

On Jan 4, 2014, at 12:10 PM, Frédéric Kayser <f.kayser@free.fr> wrote:

> Hi,
> FSE (Finite State Entropy) is a brand new entropy coder backed by Yann Collet (the guy behind LZ4) based on ANS (Asymmetric Numeral System), it could eventually be used as a interesting drop-in replacement for the Huffman encoder found in HPACK since it claims compression rates closer to arithmetic coding and faster processing than Huffman.
> More informations and code over there:
> http://fastcompression.blogspot.fr/2013/12/finite-state-entropy-new-breed-of.html
> Best regards.
> -- 
> Frédéric Kayser

Michael Sweet, Senior Printing System Engineer, PWG Chair

Received on Monday, 6 January 2014 16:33:55 UTC

This archive was generated by hypermail 2.4.0 : Friday, 17 January 2020 17:14:23 UTC