- From: Amos Jeffries <squid3@treenet.co.nz>
- Date: Wed, 16 Oct 2013 17:04:49 +1300
- To: ietf-http-wg@w3.org
On 16/10/2013 4:33 p.m., Martin Thomson wrote: > On 15 October 2013 17:21, Fred Akalin wrote: >> Just to clarify, we decided to treat header *values* as opaque bytes. Header >> names are still constrained to have only lowercase alpha characters (see >> 4.1.2 in >> http://tools.ietf.org/html/draft-ietf-httpbis-header-compression-03). > Of course, so I wonder why Herve raised #277... Still, I like the new > edits, so I'm gonna keep 'em. > I see one problem... The details about content-length specify that the response is malformed if it does not match the sum of DATA lengths. This is not a details which can be known in advance yet intermediaries are faced with the requirement of forwarding only being permitted "if the frames are forwarded without inspection of header fields". So intermediaries are thus required to wait and buffer all DATA of a response with content-length set to be received before they discover whether it is permitted to forward. This is an impossible requirement. It needs to be possible to forward responses which are not already known to be malformed, even if they turn out to be so by the end of the transaction. Amos
Received on Wednesday, 16 October 2013 04:05:17 UTC