[ih] Any suggestions for first uses of "e-mail" or "email"?

Craig Partridge craig at aland.bbn.com
Mon Aug 10 07:40:05 PDT 2015

> My understanding at the time from talking to numerous people was that =
> those who knew how knew that handling variable length addresses was not =
> really a problem and I have heard since that that many people who were =
> concerned about that have said they were mistaken at the time.

Hi John:

The team I was leading around that time (from 1992-1997) was one of the
two leading multi-gigabit router architecture teams of the time (Tony Li's
at Juniper was the other).

I heard multiple folks say variable length addresses were not a problem.
When I asked them to show me how to do it within the instruction budgets
for forwarding that multi-gigabit routers had at the time, no one could
do it.

A bit more for clarity.  There are two issues -- the route lookup and
the packet header cracking.  The route lookup was NOT the performance issue.
The issue was packet header cracking -- in 64-bit processors with limited
barrel rollers and serious hits for branches, parsing variable address
lengths and address lengths that varied from packet to packet, hurt...

In the post 2000 world, those issues are so much less severe as not to
matter.  So in retrospect, if someone thought the issue was route lookup,
then they would certainly say they were wrong.  If someone thought the
issue was header cracking, they could easily say "whoops, goofed on timeline."
But at the moment, header cracking was the issue.



More information about the Internet-history mailing list