KH> This has the disadvantage of leaving the user to wait until your
KH> render is done.
JRhine> ...we will obviously need some kind of boundary or packetized data
JRhine> scheme; no getting around that, I suspect, if we want accurate
JRhine> entity-body delimiting.
KH> I think a packet oriented approach would help. I know at one time
KH> there was talk of doing inline objects that way and interspersing the
KH> packets so that you could do what Netscape does without opening
KH> N-connections to the server. Obviously it seriously complicates the
KH> protocols though.
That is a different issue and definitely best left to HTTP NG. The original
issue was how to guarantee the client that they have received the full
entity-body, specifically how to do this for dynamic objects. I suggested
that one way to handle this was to put the entity-body on disk and then use
content-length. If you don't want to do that, I believe there are two
primary proposals: unique boundaries and packet encoding. I don't think
packet encoding was ever discussed in the context of multiplexing a
connection, as you discuss.
-- Jared_Rhine@hmc.edu / HMC / <URL:http://www.hmc.edu/~jared/home>"To hear many religious people talk, one would think God created the torso, head, legs and arms, but the devil slapped on the genitals." -- Don Schrader