(1) agree ("yuck" to content-length: -1) (2) two proposals (multipart and chunked) isn't a lot they both have their uses, so why not both (3) server push isn't 'infinite length object'. 'infinite length object' must be illegal. (4) timeouts in the middle of object transmission signal an error, not end of object (5) detecting the difference between end-of-object and 'connection gets broken' is the whole point. > However, closing the connection will still retain an important role in > the signaling of HTTP. errors yes, end-of-object no.Received on Friday, 2 February 1996 20:49:24 UTC
This archive was generated by hypermail 2.4.0 : Thursday, 2 February 2023 18:42:57 UTC