[e2e] Number of persistent connections per HTTP server?

David P. Reed dpreed at reed.com
Mon Oct 14 12:22:29 PDT 2002


At 11:04 AM 10/13/2002 -0700, Joe Touch wrote:
>TCP _assumes_ per-connection is the definition of fair. That's incorrect 
>when fair doesn't map to 'per connection'. Striping, pipelining, overlays, 
>caching, etc. all examine areas of the design space where per-connection 
>deliberately doesn't map to per-user or per-application.

Is there a document describing TCP where it says "TCP assumes fair is 
per-connection"?

I tend to think that you mean that some people in the router business (not 
the people who defined TCP) think that there is some magic about being fair 
between TCP connections.

I have said for 25 years that this is a bad idea, and in the original TCP 
design discussions we explicitly assumed that people would build 
applications that used multiple decoupled TCP connections per application 
for a number of reasons, i.e. "out of band signalling", situations where 
the relative ordering of delivery between different entities needed to be 
disentangled, etc.

This tendency of network engineers to decide that they know more about 
applications than the applications designers and users is a constant problem.





More information about the end2end-interest mailing list