[
Lists Home |
Date Index |
Thread Index
]
On 23 Jan 2002, Simon St.Laurent wrote:
> And HTTP 1.1 has lots of parts I don't consider remotely necessary for
> client-server work.
At least they're generally optional, though.
> It might be interesting to take close look at HTTP
> as a likely candidate for subsetting. GET, POST, a few headers.
>
> Won't get rid of TCP's overhead, but it could simplify some kinds of
> processing without leaping to an entirely new spec.
But consider this - HTTP over UDP:
http://research.sun.com/techrep/1999/abstract-71.html
To summarise, the client tries UDP first. If it gets no response or an
error (the server is a normal TCP-only HTTP server), it then tries again
with TCP. If the server supports UDP and the response is only a few Kb, it
replies with UDP, otherwise it just sends back the headers and a flag
telling the client to use TCP to get the body.
GET requests are perfect for UDP - just retransmit a few times if nothing
happens, since it doesn't matter if they get repeated.
ABS
--
Alaric B. Snell
http://www.alaric-snell.com/ http://RFC.net/ http://www.warhead.org.uk/
Any sufficiently advanced technology can be emulated in software
|