[arin-ppml] [Fwd: Draft Policy 2011-5: Shared Transition Spacefor IPv4 Address Extension]

[Owen DeLong]

On Feb 25, 2011, at 10:14 PM, George Bonser wrote:

>> Hi George,
>> 
>> Such an argument would be red herring (logical fallacy). I can claim
>> 128.0.0.0/8 for my own as well. When I have trouble connecting to
>> Internet hosts which overlap this range, it's nobody's fault but my
>> own.
> 
> Well, to some extent you are correct.  I mean, it certainly would be
> one's own fault but this net would be quite unlike 128/8 in that one
> would be pretty much guaranteed not to overlap with any host on the
> Internet.  And to the extent to which it is the problem only of the end
> network depends upon whether or not they are the only one a service
> provider runs into using that space.   After about the 20th one or so, I
> would guess it would become a giant pain in the service provider's hips,
> too.
> 
The difference is that this pain means that they used the space after
it was set aside for the provider to use for the conflicting purpose.
That makes it easy for the service provider to tell those users 
bluntly "You need to renumber your network, you used a range that
is reserved for our upstream connection to you."

Whereas the user would be perfectly right to argue that it is up
to the service provider to deal with their choice of RFC-1918 space.

> But if I am right, it will become obvious in a year or two.  I am not
> opposed to this, but I will have to suppress an "I told you so" moment
> when this becomes a pain in the hips being reported in NANOG every week.
> 
Providers reporting this pain to NANOG will find deaf ears IMHO.
Users do dumb things. Providers deal with this fact all the time.
Nature of the business. However, when users do dumb things
that go against rational policy, it's a lot easier to tell them so than
when they do things that are simply inconvenient to the provider,
but, technically correct (such as using whatever arbitrary RFC-1918
prefix they want).

> Where a provider will run into this is provisioning a new customer whose
> previous provider didn't use this space and is moving to a provider who
> does use the space for their WAN addressing.  It will take a couple of
> years to raise its ugly head and the way most networks operate in
> reality "it isn't a problem until it is a problem".  It won't be a
> problem for a couple of years down the road.  I'll keep an eye out for
> the stories, though, when they start to appear.
> 
Meh, of course it will happen. The thing is it will still be a much smaller
problem both in scope and difficulty.

> I two years time people will be using every scrap of v4 they can find
> and this /10 will end up being pressed into service in the most
> unexpected ways.  I hope I am wrong.
> 
I tend to doubt that. Really we're mostly talking about residential
end users here and for the most part they use whatever default
chunk of 1918 space came on their home gateway and aren't
particularly constrained by that space.

Even if they feel constrained, I'd say their far more likely to decide
to use 10.0.0.0/8 than to use some arbitrary x.192.0.0/10.

Owen