[e2e] Is a non-TCP solution dead?

Cannara cannara at attglobal.net
Fri Apr 25 16:05:18 PDT 2003


Speaking of over provisioning, others here may know more, but at the peak of
the iBubble, about 32 companies were laying fiber down the I5 corridor from
Oregon to Calif, in hopes of cashing in on all that imagined traffic that
would necessarily be fed over the NW links to/from the Far East.  Not many of
those companies left, but lots of capacity to light up the West!

On the separate thread of making traffic 'CPU friendly' (e.g., saving Acks),
just note that a full silicon TCP/IP implementation is becoming available and
a full web server inside (yes inside) an RJ45 10/100 Enet socket is now for
sale @$33US in 10k quan.  CPUs should be no one's concern.

Alex

"David P. Reed" wrote:
> 
> At 03:34 PM 4/25/2003 -0400, Nicolas Christin wrote:
> >However, this is not
> >because of poor TCP performance, but instead, because backbone links are
> >generally overprovisioned
> 
> The word is "correctly provisioned".   Slack capacity is a good thing, not
> a bad thing, if QoS matters.





More information about the end2end-interest mailing list