[arin-ppml] ULA-C

[George Bonser]

> -----Original Message-----
> From: mcr at sandelman.ca [mailto:mcr at sandelman.ca]
> Sent: Sunday, April 11, 2010 5:48 PM
> To: George Bonser
> Cc: michael.dillon at bt.com; arin-ppml at arin.net
> Subject: Re: [arin-ppml] ULA-C
> 
> A question: how many of your connections are with organizations that
> did
> not, when they first numbered their network, even know that they were
> going to connect with you?

I don't think it is a matter of knowing if they are going to connect to
*us* specifically or our knowing if we will connect to them
specifically.  It is a matter of considering if you are going to make
connections and exchange traffic in general with partners that might be
between systems not generally available to the Internet and are in
private network space.  If you decide that the probability of such a
connection is high, then it would be prudent to obtain address space
that is not likely to collide with your potential partner.  Having space
allocated from some central registry would give reasonable assurance of
that.



> I ask because, while you say:
> 
>     George> not have to worry about collisions in address space and
>     George> would probably be worth the expense for us provided it was
>     George> for our partners as well.
> 
> that it's worth the money, that's only with the knowledge that you'd
be
> connecting.

Sometimes "peace of mind" is worth a bit of money.  If the network can
be rolled out with the reasonable expectation that there will not be a
collision, ever, with a potential partner, then it is well worth $1200.
How much time would be spent in engineering an alternative if there is a
collision?  Will it take more than 3 hours to figure out and implement?
If it does, then the money was well spent.  $1200 is practically free.

> What if you did not know, a-priori, that you were going to connect?
> How cheap would ULA-C have to be to just "get some" regardless, vs how
> much expense do you think renumbering is?

Again it is a matter of knowing that you are going to have connections
with people.  You don't know in advance what their address space is, you
don't know in advance even who exactly you are going to connect to.  To
absolutely ensure that you will never have that problem is a good
architectural decision.  Not doing so is probably fine for a "garage"
operation that lives by the notion that "it isn't a problem until it is
a problem" but that isn't the best way to do things, in my opinion.

I would rather do it so the problem will never occur in the first place.