US CODE: Title 15, Chapter 1, Section 2.

Howard C. Berkowitz hcb at CLARK.NET
Sun Feb 2 20:06:24 EST 1997


Randy,

I fully agree many of the issues here are more NANOG/IEPG. Unfortunately,
as you suggest, a significant number of participants here either don't know
of those groups, or don't understand the relationship between ARIN and
NANOG/IEPG. I'm not sure how to refocus this list onto address registration
(ignoring ASN's).  People do bring up legitimate concerns, and perhaps
there needs to be more of a forum for putting them in context.

At 3:30 PM -0800 2/2/97, Randy Bush wrote:
>I share your concerns.  I also share your questioning if this is the place
>to solve them.  Smells a bit NANOGish to me.
>
>> To my mind this obligation isn't so much a list of specific do's and
>> don'ts (although over time those might evolve), but rather as a general
>> mutual obligation to try to honor the grants.  By this, I mean that a
>> member is obligated to be reasonable (I know this is vague) about routing
>> information for ARIN grants.
>
>While suspect it may not be wise to burden ARIN with this mandate, it has
>enough problems already, could something be done to communicate this need
>to the culture as a whole?
>
>Observe that this is current accepted practice for reasonably aggregated
>routing.  But, as has been so well demonstrated on this list, what is an
>accepted part of the culture today seems not to be understood by the
>massive influx of new folk.  We should probably be less surprised than I,
>for one, am.
>
>> What I'm fumbling with (and fumbling it is) is the example that you
>> brought up -- the medical group that justifies a dual-homed, provider
>> independent class C.
>

It's a deliberately nasty example, and even then, probably better than in
the real world.  I stipulated the medical group had thorough exterior
routing competence.  That isn't as likely as we might like.

>What if NSI or ARIN will not allocate less than a /19, and they keep
>qualifications for space about the same as they are now?  I.e. the medical
>group has no alternative but to take its /24 from one of its providers.
>The question becomes whether their other providers will accept that /24
>from the medial group and propagate it.  And will any of their peers
>listen to a /24?

Without getting into any issues of what might give an appearance of
collusion, I think the long-term solution to this sort of thing will be
specific, fairly routine, inter-provider backup arrangements.  My intuition
is that aggregation is probably best served if the medical group gets a /24
assigned from Provider A space, with prior agreement from Provider B that
Provider B will advertise the specific hole from Provider A space, if and
only if the customer-to-A link is lost and the customer intelligently
starts advertising the /24 to Provider B.

Not necessarily pretty, but winds up with only one /24 being advertised
occasionally.  Of course, that /24 needs to be withdrawn when the primary
link comes back up, so some level of dampening probably is necessary.

One would hope that with a sufficient number of bilateral provider
agreements like this, the workload would tend to equalize.

Now -- this is a real world problem.  I haven't even scratched issues like
attempts to load balance, having >2 providers, etc.  I'm sure that if this
were discussed in detail, there would be many more complexities.

In all sincerity, for those of you who think ARIN is being anticompetitive
because it discourages giving out small PI blocks, please read the
preceding four paragraphs.  I have not tried to be obscure, but have not
written a lengthy tutorial either.  If you are uncomfortable pointing out
the strengths and weaknesses of this off-the-top-of-my-head example, I
respectfully suggest you have some homework to do before assuming that
other than technical factors are involved in the allocation & routing
policies here.
>
>Before folk who do not run default-free routers today go ballistic, note
>that this issue is current today, sans ARIN and independent of the
>InterNIC.  If someone bludgeons NSI into giving them a /24 out of 208 or
>wherever Kim's allocating this week, it is becoming less and less routable
>as more and more of us unstall filters.
>
>So if the InterNIC allows the medical group to bludgeon them out of a /24,
>they are really not doing the medical group a favor, in fact, it is a
>disfavor.  I suggest that ARIN consider not doing so.

In general, I agree.

>
>But, as you suggest, much of this is NANOG and IEPG fodder, as it is an
>inter-pprovider matter, and neither InterNIC nor ARIN can dictate to the
>providers.
>
>randy






More information about the Naipr mailing list