On Tue, 18 Apr 2006 14:46:15 -0400, "Theodore Ts'o" <[EMAIL PROTECTED]> wrote:
> On Tue, Apr 18, 2006 at 11:42:27AM -0400, Keith Moore wrote:
> > > It smells remarkably like pathalias to me ;-)
> >
> > except that I'm not proposing that border routers do source routing,
> > just that they map from PI identifiers to PA locators and prepend a
> > header that causes the payload to be routed to the locator.
>
> And that sounds exactly like pathalias and what the Usenet/SMTP smart
> routers did, yes? Hopefully though they wouldn't require as much CPU
> power to calculate as pathalias database required --- as I recall Erik
> Fair at Apple used a Cray for that purpose, because he didn't like
> waiting.... :-)
>
Peter Honeyman optimized my original algorithm considerably; there's even
a paper on it on my web page. There were two more fundamental problems
that neither of us every really solved.
One is relatively easy today: we didn't have good maps of connectivity for
each site. Today, though, we have routing protocols that are supposed to
disseminiate such data. Even now, though, there's a tension between
physical connectivity and policy; we ran into that, too. The harder
problem, though was metrics. Simple hop count gave bad results; bandwidth
varied. (That said, the probability of uucp silently dropping the message
was high enough that maybe we should have stayed with just that as a
metric....)
--Steven M. Bellovin, http://www.cs.columbia.edu/~smb
_______________________________________________
Ietf mailing list
[email protected]
https://www1.ietf.org/mailman/listinfo/ietf
