current pTLA list - 28Aug97
Fri, 29 Aug 1997 16:05:19 -0700 (PDT)
>>>>> "Craig" == Craig Metz <email@example.com> writes:
Craig> In message <199708291834.LAA08399@pedrom-ultra.cisco.com>,
Craig> you write:
>> I'd like to carve addresses from a separate TLA for
>> interconnects (i.e. a TLA not associated with any site). It
>> would be nice if somebody will step in and administer such a
>> thing... if not i guess i'll have to volunteer.
Craig> I can do this if we decide that it's a good thing to
Craig> However, I'm not sure that it's a good thing to
Craig> do. The question here is whether the addresses associated
Craig> with tunnels should be globally reachable or not.
Craig> I would
Craig> suggest that they should be, in which case they should come
Craig> out of the prefix of one of the two sides of the tunnel.
It has been done this way until now and it works... The only pontential
problem is that when one site renumbers (and i really don't want to
discuss if renumbering is a good idea) you "loose" the prefix.
Craig> This would all be *a lot* simpler IMO if
Craig> implementations had a mode where the global and/or
Link locals cannot be used.
Craig> off some other interface could be stolen for BGP
Craig> neighbor purposes instead of treating a tunnel as a
Craig> bi-directional point-to-point link with both global and
Craig> link local addresses.
That can be done with global addresses. There are however good reasons to do
BGP over bi-directional point-to-point links, simplicity being one of them.