- From: Yves Lafon <ylafon@w3.org>
- Date: Thu, 8 Dec 2011 09:48:55 -0500 (EST)
- To: Mark Nottingham <mnot@mnot.net>
- cc: "William A. Rowe Jr." <wrowe@rowe-clan.net>, ietf-http-wg@w3.org
- Message-ID: <alpine.DEB.1.10.1112080942120.26187@wnl.j3.bet>
On Wed, 7 Dec 2011, Mark Nottingham wrote: > > On 24/11/2011, at 3:52 PM, William A. Rowe Jr. wrote: > >> It's easier to say servers are always permitted to coalesce responses in a >> manner that makes delivery more efficient. I believe this needs to include >> sequencing them in serial order as mentioned in... > > Reading this thread, I'm inclined to agree; rather than being too > specific, we could note the security issues, as well as the potential > impact on clients. > > How about adding a paragraph to p5 5.4.2: > > """ Servers are not required to return the exact range requested in a > partial response, and MAY coalesce several ranges into a single > response, to make delivery more efficient. Clients SHOULD NOT depend > upon the requested ranges being returned as specified in a partial > response. This includes the size of the ranges, their offsets, and their > ordering in the response. > """ There is already a paragraph about that issue in the security section, but yes, something needs to be added to explicitely allow servers to coalesce overlapping ranges. Also note that in 5.2: << When an HTTP message includes the content of multiple ranges (for example, a response to a request for multiple non-overlapping ranges), these are transmitted as a multipart message. The multipart media type used for this purpose is "multipart/byteranges" as defined in Appendix A. A response to a request for a single range MUST NOT be sent using the multipart/byteranges media type. A response to a request for multiple ranges, whose result is a single range, MAY be sent as a multipart/byteranges media type with one part. A client that cannot decode a multipart/byteranges message MUST NOT ask for multiple ranges in a single request. >> Which is implicitely authorizing coalescing ranges and hinting that overlapping ranges in multipart/byteranges should not be overlapping. But it's better to explicitely say it. (There are also some examples missing) I'll send a diff proposal soon. -- Baroula que barouleras, au ti騏 toujou t'entourneras. ~~Yves
Received on Thursday, 8 December 2011 14:48:58 UTC