ARGH!!!! On Fri, 6 Sep 1996, Peter Lothberg wrote:

...

Second: allowing such a customer, or an NSP, to attach web services directly to the FDDI ring at the NAP.

Peter: If I had thought through the topology of the situation I too blithly described in the above sentence, the pieces of equipment involved, and what bits were flowing where, even at my not really sophisticated level of knowledge, i might have seen the problem that placing a web server with nothing between it and the gigaswitch FDDI port would have involved. Instead, feeling far too confident that I remembered a 14 hour old phone conversation with stephen stuart correctly, I wrote what i thought I had heard. I am fully aware now that I heard mistakenly. But I also had no earthly idea what a cow pie I had stepped into. I would be happy to let this die. But since I was not and am not trying to cause problems for the Palo Alto digital people I have no choice but to answer it - for you write as though he (stuart-dec-paix) had not corrected my error - something that he has done.

This is a security problem, if there is no switch in the middle and each host are induvidually attached to the switch.

Next problem is that a host needs to knew what router to send a packet to for a particular destination, so either it points default at one of the NAP routers, and packets traverse the NAP twice, or the host impleements BGP and has a full set of routes.

So host at a the NAP media should be 'strongly not recomended'.

Thank you for a good explanation of some of the major reasons why such a topology would be ill-advised. ;-)

An interesting scenario is, a router with two FDDI interfaces, one to the host and one to the NAP. It now comes down to if it's worth the real_estate to have the host there.

What you have just described, is what, if I now understand things correctly, exists at the Palo Alto Internet exchange.

--Peter

Gordon Cook