[j-nsp] ISIS adjacency on the GigE interface

Kevin Oberman oberman at es.net
Mon Jun 13 16:11:39 EDT 2005


> From: Dave McGaugh <dmcgaugh at cac.washington.edu>
> Date: Mon, 13 Jun 2005 12:48:24 -0700
> Sender: juniper-nsp-bounces at puck.nether.net
> 
> I'm not an ISO expert, but here goes :)
> You might try changing the AFI from 99 to 49. Some docs I've read  
> seem to suggest that the AFI is, among other things responsible for  
> defining the addressing format. I'm not sure whether JunOS enforces  
> such a thing, but if your addressing structure conflicts with that  
> which is defined for AFI 99, it could create problems..
> 

Dave,

You have it right. AFI is the Authority and Format Identifier. 49 is the
"binary local" space and the format is 20 octets. It should be used if
you do not have a ANSI assigned NSAP. 99 is probably bogus, but I can't
find my OLD OSI/GOSIP documentation to confirm it.

>From RFC137:

   The only other defined Authority and Format Indicator (AFI) which
   leaves sufficient space for both an IPv6 address and TCP port number
   is the binary local AFI (49). 

So, if you don't have an official NSAP (probably with an AFI of 47), 49
is the one to use. It is 20 octets ling and normally in the format is
written as:
49:xxxx.xxxx.xxxx.xxxx.xxxx.xxxx.xxxx.xxxx.xxxx.00

Note the final 00. This is the selector byte (don't worry what that is)
and should always be 00. It is STRONGLY recommended that something
meaningful be encoded in the 18 available octets. You can fit an IPv6
address in there just fine or do something else. Just as ling as it is
unique in your ISIS fabric.
-- 
R. Kevin Oberman, Network Engineer
Energy Sciences Network (ESnet)
Ernest O. Lawrence Berkeley National Laboratory (Berkeley Lab)
E-mail: oberman at es.net			Phone: +1 510 486-8634


More information about the juniper-nsp mailing list