Lots of non-existant networks

Jake Khuon khuon at Merit.Net
Thu Jun 26 19:35:16 UTC 1997

### On Thu, 26 Jun 97 12:02 PDT, randy at psg.com (Randy Bush) wrote to Abha
### Ahuja <ahuja at Merit.Net> concerning "Re: Lots of non-existant networks ":

RB> > http://www.rsng.net/pair/
RB> I am confused as usual.  E.g.
RB>     http://www.rsng.net/rs-views/mae-west/AS2914/grey.html
RB> shows
RB>     GREY (4553) 2914 IGP

This is fine.  As a matter of fact, all AS2914 routes in the RSes should be
GREY because you have specified to not have any of your routes redistributed
by the RSes.  GREY does not necessarily indicate "bad".  It just denotes a
route that the RS knows about but which has no policy for because the
evaluation engine returned a null.  The order by which policy filters for
each peer of the RS is evaluated as follows.

[1] Route Server Peering Request (extended inet-rtr)
[2] aut-num registered in the IRR
[3] route registered in the IRR

In the case of [1] and [2], pairwise reciprocal policy must exist otherwise
a null is returned.

RB> Yet the registered policy is
RB>     route:
RB>     descr:       RG-TIG-NIC-0
RB>     origin:      AS2914
RB>     remarks:     this is non-portable space, no shorter prefixes allowed
RB>     notify:      rw at rg.net
RB>     mnt-by:      MAINT-RGNET
RB>     changed:     randy at psg.com 960829
RB>     source:      RADB

Actually that's the registered route.  Registered policy (aut-num) is also
fine, however, Route Server specific policy (via the import()/export()
statements in your peering session as specified via the inet-rtr object you
submitted to us) evaluates out to a null list, thus we do not redistribute
your routes or distribute routes to you.  In the future, we will include
"shades" of colors (by comparing the data we already have with additional
datasets generated during the policy evaluation phase of the RSes'
reconfigs) to more greatly explain and distinguish not just the status but
the reason for the state of each route.

RB> So what is inconisitent about this?

The fault here is ours.  Currently we state in our pages that GREY routes
are routes that do not match prescribed policy in the IRR.  However, in
actuality, GREY routes are routes that _either_ do not match policy specific
to peering with the RSes _or_ policy registered in the IRR.  In the very
short future, policy for route servers via the optional extended attributes
to the inet-rtr will be merged into the RADB/IRR so the original statement
will be true.

The list of GREY routes for your network actually is proof that things are
consistant with what you want.  I suppose in some ways you can view it as an
inconsistancy if you expected the RSes to pass your routes on but since you
don't then there is no inconsistancy.

/*===================[ Jake Khuon <khuon at Merit.Net> ]======================+
 | Systems Research Programmer, IE Group       /| /|[~|)|~|~ N E T W O R K |
 | VOX: (313) 763-4907   FAX: (313) 747-3185  / |/ |[_|\| |   Incorporated |
 +==[ Suite C2122, Bldg. 1  4251 Plymouth Rd.  Ann Arbor, MI 48105-2785 ]==*/

More information about the NANOG mailing list