- From: Jeff Pinner <jpinner@twitter.com>
- Date: Sun, 16 Jun 2013 11:52:08 -0700
- To: James M Snell <jasnell@gmail.com>
- Cc: HTTP Working Group <ietf-http-wg@w3.org>
- Message-ID: <CA+pLO_hJk2wVEUu3AsK+QFocA3s0Y=2=+-cG5O4HTmg6yV1rog@mail.gmail.com>
Another point requiring clarification: The draft lists tables with pre-defined headers for "requests" and "responses." I am presuming that the intention was, for example, that a server would initialize its sending header table with the "response" table listed in appendix A.2. My question is, should the server also use this table when compressing the PUSH_PROMISE header block which contains "request" headers? On Sun, Jun 16, 2013 at 8:13 AM, James M Snell <jasnell@gmail.com> wrote: > It is currently unspecified. There are two ways of doing multiple values > in the current draft that may change a bit once we get into the type > codecs. Currently you can either null separate the instances in a single > header value, or you can use separate opcodes for each (treat them as > separate headers). The latter can be far more efficient in the encoding. > For now, I'd recommend the separate opcode approach and revisit it when we > discuss type codecs in more detail. > On Jun 16, 2013 7:48 AM, "Jeff Pinner" <jpinner@twitter.com> wrote: > >> The SPDY spec defined how multiple headers with the same name were to be >> encoded (as one header with the value-field containing multiple >> null-separated entries). >> >> Is this the expected encoding >> for draft-ruellan-http-header-compression-00.txt? >> >
Received on Sunday, 16 June 2013 18:52:35 UTC