[e2e] Open the floodgate

[Eric A. Hall]

On 4/24/2004 2:53 AM, Sam Manthorpe wrote:

> This is probably a dumb question, since I've been out of networking for
> quite a while, but doesn't the TCP/IP stack *have* a network-layer 
> congestion amerlioration mechanism: ICMP source quench.  It seemed to
> me that this never got used because it was ill defined as to what the
> ends should do when they receive such a message, plus VJ's congestion 
> avoidance algorithm eliminated its necessity for 99% of traffic.

The surviving argument against it is that routers couldn't handle the
slow-path processing they typically performed.

Thus the final argument against making new routers generate ICMP
efficiently was that the old routers (the ones being replaced with new
routers) couldn't generate ICMP effeciently.

http://www.postel.org/pipermail/end2end-interest/2001-February/000113.html

-- 
Eric A. Hall                                        http://www.ehsco.com/
Internet Core Protocols          http://www.oreilly.com/catalog/coreprot/