Re: Protocol Benchmarking (with Accept examples - long)

Jon P. Knight ([email protected])
Thu, 3 Feb 1994 09:04:37 +0000 (GMT)

On Wed, 2 Feb 1994, Lou Montulli wrote:
>
> If the host sent back the type immediately after the client sends
> the get, the client could check it's list of accepts to see if
> it's an acceptable type, for instance "image/jpeg".
> If the type is acceptable, the client responds, "OK
> send me the data", otherwise the client says "I don't understand
> image/jpeg" but I do understand "image/gif and image/x-xbm".
> If the server can deliver those types then he sends the data,
> if not then the server may attempt a different sub-group, for
> instance "application/x-mac-draw". In each case the client
> would only send the accept header of the specified sub-group,
> thereby saving the broadcast of a very large group of accepts.
>

Correct me if I'm wrong but currently we send Accepts in a single outward
call and get some form of result (either the document or an error)
returned from the server. With your scheme we send a request, receive a
list of possible types, send back a request for types the client can
handle and then get the reponse from the server. Therefore we've gone
from one round trip delay to two. Thus to save transmitting a few bytes
we take an increased latency hit. Right? When some of the network links
have RTDs in the thousands of ms from where I'm sitting and yet we have a
nice fat pipe to the Internet, I think we'd rather waste a few bytes on
small documents. Latency is Your Enemy(tm) in WAN based distributed systems.

Just IMHO.

Jon

-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
Jon Knight, Research Student in High Performance Networking and Distributed
Systems in the Department of _Computer_Studies_ at Loughborough University.
* Its not how big your share is, its how much you share that's important. *