Re: Last Call for draft-ietf-rolc-apr-00.txt

[Curtis Villamizar <[e-mail]>]

In message <[e-mail]>om>, Yakov Rekhter writes:
> Curtis,
> 
> So, to clarify the (apparent) confusion:
> 
>    APR != whole NBMA network.  Whole NBMA network may contain more than one A
> PR.
> 
> Yakov.


If we keep APR, then the defintion is simple.  APR is a prefix that
falls entirely within an NBMA.

You can describe the "one hop off the NBMA" problem as:

  Where direct reachability is being determined by a preconfigured set
  of prefixes, a prefix served by a router on the NBMA, for which direct
  connectivity to that router is sometimes desired, must be configured
  in the set of prefixes.  Though this is technically not an APR, this
  prefix may be configured as a subset of an APR to accomplish the
  same result.

The above describes a violation of the purity of an APR.  It may
become common practice to avoid configuring additional prefixes at the
expense of this architecural purity.

I would certainly rather define APR to describe the role in the
topology or the role in the hack to simplify configuration.

My preference is not to introduce a new term at all.  If the use of
address prefix to determine direct reachability is a key point of this
draft, then it is a weakness of the draft, in that it is concentrating
on a single very static means of determining if a destination is
directly reachable.  There is only a need for the term APR if the
draft dwells on this very poorly scalable technique.

[ aside: I thought large scalability was a key goal of the ROLC WG.
In this preoccupation with use of APR to determine reachability and in
the current NHRP, that goal seems to have been lost completely. ]

Curtis