The Routing Over Large Clouds Mailing List Archive by date

[Masataka Ohta <mohta@necom830.hpcl.titech.ac.jp>]

Andy;

> As has been noted on the list, there has been a lot of "interest" (a
> polite word for "hype") lately in the industry with regard to cell
> switching routers.

I'll happily attack any hype on CSRs. So, please give me a specific
references of the hypes.

> However, I (as an equipment purchaser) would think twice before
> purchasing any router, cell switching or otherwise, that doesn't meet
> at least the most important requirements in RFC 1812 (Requirements for
> IP Version 4 Routers).

Hmmm, there seems to be a hype AGAINST CSRs.

Before thinking twice, you should read RFC 1812, 1122 and 791
ten times.

> High on the list is the requirement to
> decrement TTL and not forward packets with a TTL of zero.

Do you insists on the precise wordings of Internet Standards, not
it's intension?

OK. Fine. While usual engineers don't mind wording so much,
everything should by justified by you to protect NHRP.

BUT, if you have precisely read RFC 1812, you could have noticed
it say things like "forward datagrams" or "IP datagram was queued
for forwarding".

And, CSRs, when forwarding datagrams, do decrement TTLs. When
forwarding cells, they may not.

RFC 791 is a little more explicit that it says:

    This field must be decreased at each point that the internet header
                                 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
    is processed to reflect the time spent processing the datagram.
    ^^^^^^^^^^^^

That is, bridges and cell relays do not have to reduce IP TTL.

> Ipsilon,
> for one, recognized this requirement

I have been aware of this TTL issue from day one of CSR.

Do you want to see my first Internet Draft on CSRs?

> and found one way to work around
> it while switching at the cell layer; I'm sure that there are other
> solutions as well.

Really?

That could be an additional protection agaist the weakness of ATM
(including NHRP). But, unlike on usual medium, it is not so
important, because, even if there is a cell forwarding loop,
policing will prevent excessive consumption of bandwidth.

> Given the existence of CSRs, does the need for NHRP go away?

That's not a logic of anyone.

NHRP must go away because it does not scale.

> Cheers,
> Andy

My cheers.

						Masataka Ohta