We need to resolve the issue of content-length and cascaded transfer encodings quickly. We need two interoperable implementations of this 'feature', even if it is implicit in the spec, or else we should restrict the allowed set. I don't think we should explore too many degrees of freedom. * chunked is only allowed once, as the last transfer encoding applied. * before chunked is applied, only one T-E should be sent, but recipients should accept all combinations (as long as there are no duplicates). * no T-E's other than 'chunked' may be applied to multipart/ content-types, but T-Es are allowed within a multipart type (e.g., multipart/byte-ranges, multipart/form-data). The entire transmission is required to be either with content-length or else self-delimited where multipart is the only self-delimited media type, but chunked, gzip are self-delimited T-Es. -- http://www.parc.xerox.com/masinterReceived on Wednesday, 17 December 1997 12:43:04 EST
This archive was generated by hypermail pre-2.1.9 : Wednesday, 24 September 2003 06:33:05 EDT