Hi,

A couple of issues to think about.

First: the NLPID as defined uses the low 4 bits as flags; this means
we are using a block of 16 NLPID values in effect. This is NG. The flags
need to be moved elsewhere. There is also a need for another flag for
Error-report Suppression.

---

TTLs: some feedback and further thinking leads me to believe that
scaling the TTL when translating datagrams is not such a good idea
(although it must be done when going to and from IPX). It would be
better to keep it to hop count.

We still want a nominal time associated with it, for when a router
hangs onto a datagram for a longer than usual time (typically waiting
for some kind of subnetwork setup). Probably 1/10 second is good; it
must be less than IP's (1 sec) and less than CLNP (500 msec)

It also seems that 8 bits is really sufficient. It is a log function
of network size. If a net with 1 billion hosts has a topological diameter
of (say) 30, then a diameter of 250 with the same branchiness implies
something like 10^82 hosts. (There are something like 10^78 neutrons
in the observable universe if I recall correctly. :-)

Maybe it would become less branchy; but recent trends are in the other
direction. Subnetwork technologies like ATM will push the internetwork-
layer diameter down even further.

(Does anyone have a candidate for longest path in the existing Internet?

---
Taken together, this means the first 32 bits of the CATNIP header
would be:  NLPID (70), header length, flags, TTL (each one byte).

Comments anyone?

Rob