- From: Martin Thomson <martin.thomson@gmail.com>
- Date: Thu, 3 Apr 2014 23:45:54 -0700
- To: Matthew Kerwin <matthew@kerwin.net.au>
- Cc: "ietf-http-wg@w3.org" <ietf-http-wg@w3.org>
On 3 April 2014 22:21, Matthew Kerwin <matthew@kerwin.net.au> wrote: > Yes, multiple values. I didn't think it would be necessary to tell > implementors how to process or store the values. I guess I could add > something to the following effect: > > Have a table of encodings you are able to transmit, with ranks initialised > to 0. When you receive a this setting, if it's for one of the encodings in > your table, update the rank in the table from the frame; if it's not, ignore > it. You can use your table to help inform your decision of what encoding, if > any, to apply to any DATA frames. > > Does that not count as an implementation detail? That's not the point. The point is that an endpoint needs to be able to limit the state commitment it makes. Based on this, the commitment could be as much as 256K for these settings.
Received on Friday, 4 April 2014 06:46:25 UTC