Huffman [was: Re: h2 header field names]

--------
In message <CABkgnnVEyR2BV7gf_BANXN=cF3UbdV=vwEP_vij6bRCgBubVMQ@mail.gmail.com>
, Martin Thomson writes:
>On 5 September 2014 09:25, Poul-Henning Kamp <phk@phk.freebsd.dk> wrote:
>> Can I propose that the origin of the data and the procedure used
>> to derive the table be added to the draft as an appendix.
>
>You certainly can propose pretty much anything you like.  I personally
>don't see the point.
>
>I'd be interested in learning why you might think that tampering would
>produce a benefit for anyone.

There are no accusations of tampering anywhere I can see, and it's
not about what I think.

This table is a key element in HPACK and therefore their provenance
should be documented, if nothing else for historys sake.

I found this email:

	http://lists.w3.org/Archives/Public/ietf-http-wg/2014AprJun/1196.html

Which seems to be for a check-run to see if the current able should
be updated ?

It doesn't answer my question if HTTP/1 headers which cannot ever
be sent through HTTP/2 were filtered out before estimating the PDFs.

I'm thinking specifically of:

	"HTTP/1.1" (HTTP-Version in the status line)
	Reason-Phrases in the status line
	Connection: close
	Connection: keepalive
	Transfer-Encoding: chunked

None of these will pass through HTTP/2.0 and should therefore not
be used as training material for HPACK

Does anybody know if these headers were removed from the data-set ?

-- 
Poul-Henning Kamp       | UNIX since Zilog Zeus 3.20
phk@FreeBSD.ORG         | TCP/IP since RFC 956
FreeBSD committer       | BSD since 4.3-tahoe    
Never attribute to malice what can adequately be explained by incompetence.

Received on Friday, 5 September 2014 17:06:10 UTC