[arin-ppml] Multihomed Microallocations

Ted Mittelstaedt tedm at ipinc.net
Tue Aug 4 18:32:07 EDT 2009

Kevin Kargel wrote:
>> Why does the small user and experimenter WANT PI space in the first
>> place?  Well, their multihoming I hear you say.  Baloney.  You do not
>> need PI space to multihome.  You can advertise an assignment from an
>> upstream just fine.  So what that there's a covering route out there,
>> you still have your redundancy.
> Gee, last week when I suggested doing it that way people jumped all over me
> telling me how that was broken and next to worthless..  I do agree with them
> to a large degree based on what happens if I give IP space to a multi-home
> customer, announce his network and the aggregate, another ISP announces only
> the small network, and then my connection to the customer goes down.  
> The only way (to my understanding and as I was told last week) to get a
> fully functional redundancy model without LOTS of silly router tricks is for
> the customer to have their own ASN.

Agreed.  I assumed an org (or person) would have their own ASN when
they were advertising a block allocated to them from their upstream.

That is how we did it for 8 years and it worked fine for us.

>  When they have their own ASN I could
> de-aggregate what I allocate to them, but once they have their own ASN
> anyway then the 'might as well have their own netblock' comes in to play. 

No it doesn't.  As I said, we did it this way for 8 years before getting
portable space from ARIN.  If your a new ISP starting out how on earth
can you predict if your business will be a success or failure?  How
can you predict your future customer base until you go out and get
those customers?  You need to sell your services first, then worry about
getting portable numbers later, once you know what your customer base
will be.

The same things apply to PI space that's not reassignable.  If your
an org that starts out multihoming or some such, then how are you
going to know that the multihoming is even going to do what you think
it will do or help your org out at all, until you do it?   How do
you know your going to want to pay the extra money for it on an
ongoing basis?  Maybe your a small org with a connection that goes
down 20% of the time so you multihome, and discover that the new
connection goes down 1% of the time, while the original is still going
down 20% of the time.  Maybe you decide later that paying double
for multiple connections is stupid when you can save the money and
just keep the connection with 1% outage.

There's good reasons to start out with space assigned from an upstream.

There's also good reasons for getting your own space once the
provider-assigned space has proven itself and you know that you
have a need for the portable space.

>> When you boil it down it really comes to 2 reasons:
>> a) The small operator wants to be able to tell an upstream to kiss
>> off if they get pissy with him, without renumbering.  That's lock-in.
> This is not trivial.  Allowing provider independence is a good thing. I
> enjoy it and I expect others do also. 
> If we are trying to write ARIN policy to support an addictive business plan
> I would call that a conflict of interest which should be vigorously avoided.

I wouldn't tell people to do things I wouldn't do myself.  We ran for
EIGHT YEARS with upstream-assigned space.  Don't lecture me with scare
stories about how terrible this is, I know most of that is hogwash.  It
seriously is just not that big of a deal as made out to be for small
allocations.  And we had a LOT of space, much more than a few /24's,
tied up this way.

Most of the gasoline consumed in US cars comes from the MidEast and
if that's not an addictive business plan with numerous conflicts of
interest I don't know what is.  Yet it's tolerated and even pushed in
some quarters.  Even today there's still people who would rather see
more oil wells than solar-power development.

"addictive resources" are present in many industries, the Internet is
not special at all in this regard.

>> b) The small operator cannot get even a small amount of space (like
>> a /24) from an upstream because that upstream is severely constrained on
>> IP.
> While I don't think that is typically the case now, I suspect it is
> possible, and in the future will be likely.

I agree that it will be an issue in the future.  It will be an issue for
everyone, of course.  Some sooner than later.  Likely, the small orgs
will be affected earlier.

> c) They are interested, they want to educate themselves, experiment and do
> it like the big people.  This is ok in my book and qualifies as a
> justification.  Remember that there are costs and red tape hurdles to go
> through that will limit this type of exercise.  Everybody has to start some
> where, and this looks like a good starting point to me.  

What about the people who want to educate themselves 5 years from now
when the only way to get more IPv4 is by buying it for lots of money in
large chunks?  Sooner or later the Internet is going to have to fish or
cut bait with IPv6.

>> Neither reason is really valid now.  Renumbering a /24 is childs play,
>> if an "experimenter" or "small user" cannot manage to renumber a /24
>> they are incompetent and shouldn't be fooling around in this to begin
>> with.  So what it's inconvenient - their desire for less inconvenience
>> is at the expense of the entire Internet.  Thus, for a /24, lock-in is
>> a myth.
>> As for the upstream being constrained, well I don't see ARIN denying
>> IPv4 requests right now.  Sure they will in the future - but there will
>> be plenty of small /24's that are currently abandoned that ARIN will end
>> up taking back as a result of recovery efforts.  If an upstream is
>> claiming they are IP constrained that is a flat out lie.  Send them to
>> the RIR for more numbering.
>> Now, maybe in the future this will change and little allocations won't
>> be available anymore.  So people are wanting to get their
>> microallocations now, before the land rush.  OK, fine, no problem there.
>>   But eventually no matter what policy is done, there won't be any
>> more allocations of any kind.
>> You might be throwing a lifeline to the small users and experimenters
>> right now, but only for the next few years crop of 'em.  After all,
>> what do you think becomes of experimenters when their IPv4 disappears?
>> Those that have brains and courage come out all right; those that
>> haven't are winnowed out.
> Well, if policy for the remainder of IPv4 life is only the next few years
> and doesn't matter then we should all save ourselves a bunch of work and
> quit wasting our time discussing IPv4 policy.  I don't buy this line of
> thinking.

Neither do I, I do not agree that once IPv4 runs out that there will be
no need of IPv4 policy, in fact, I think the reverse will be the case.

> I really don't see this becoming a big burden on the backbone, and I see no
> need to slam the door on the few that may want to do it and will be willing
> to pay for it.

I don't either, but you keep flipping back and forth on your point, Kevin.

This terminology for example:

"...Working to exclude the experimenters and small users will ultimately 
prevent good that could be done for the community....Rather than 
fostering an elitist attitude and trying to keep anyone who is not as 
big as us out..."

contrasts with this:

"...slam the door on the few that may want to do it..."

The FEW.  Kevin, you said few, not me.  Your assuming, perhaps 
unconsciously, that there's only a small number that want to do this. 
Yet many times you make claims about the large number of people and orgs 
that want to do this.

I know your trying to be fair but I think that you have already
unconsciously accepted the premise that there needs to be a limit,
a "speedbump" as it were to just handing out small allocations
to anyone who asks.  You know, in your heart, that there is logic

I am just saying here that there's good reasons that it has been
made difficult to get this space.  Now perhaps some of those are
not as valid anymore - routers today are more powerful and can hold
more routes, that kind of thing.  As well as the upcoming end of
IPv4.  Those are changed premises.  But, not ALL of the premises
originally used as reason to make that entry difficult have disappeared.
I would therefore go very, very cautiously here.


More information about the ARIN-PPML mailing list