Re: h2 header field names

--------
In message <4CFA7625-24C4-49CB-BCED-32598C181ACC@mnot.net>, Mark Nottingham wri
tes:

>Because chunk boundaries have no semantics in HTTP/2, whilst headers do. 

The argument for not tightening up header definitionis that "weird
stuff" have semantics in practical HTTP/1.1 so we can't do away with
it, in case we need to tunnel it.

Chunk boundaries also have semantics in practical HTTP/1.1 and we
just did away with them, even for tunneled HTTP/1.1.

What is the difference ?

There is a valid architectural point to decide ASCII vs. UTF-8, and
we can and should debate that.

But I havn't seen *anybody* say that need to be able to put NUL,
STX or ANSI-escape sequences in HTTTP headers, so I don't understand
why can't we outlaw them in HTTP/2.0, even if we don't settle the
ASCII/UTF-8 question yet ?

IMO nothing *in* the headers should contain 0x00-0x1f or 0x7f.

What makes that decision impossible ?

-- 
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 Thursday, 4 September 2014 12:02:45 UTC