Re: SIP Addressing Limitations

["Paul Francis (formerly Paul Tsuchiya" <[e-mail]>]

>  
>  > >From where would you allocate space for a provider that
>  > doesn't fall under one of the clusters?  For instance,
>  > a provider that has direct customers in the Northeastern
>  > USA and the far east (Singapore and Korea, say)?
>  >
>  It is becoming apparent that you haven't really read the proposal.  Try
>  "Constraints", on page 2.

Yes, I saw that section.  I just wanted to make sure that
I understood your intent to leave only 1/8 of the address
space for everything that didn't fit your model.....

(I understand that, in theory, the IPAE half will someday
be given back, when all the IP hosts go away, but I think
that won't happen for a very very long time....)

>  
>  I am unable to answer the question, since you didn't provide enough
>  information.
>  
>  How many customers?  10,000,000?  10,000?  10?
>  

Well, let's say it is 50 today but grows to 500,000 in 10 years.
It would be nice if it didn't matter how many customers it had.....

>  
>  Finally, I don't think the IANA would spend much time considering such a
>  provider.  Since it has a private trans-continental link, and a private
>  trans-oceanic link (or a very very long trans-oceanic link around the
>  Cape, or a very long multi-satellite path), it will be priced far in
>  excess of the costs of its competitors.  It will not receive funding
>  from any but the most foolish investors, and have none but the most
>  foolish customers.  It will fail in the marketplace.
>  
>  Why are we wasting our time with theoretical examples of providers that
>  aren't connected to the internet?

Well, I will confess that I baited you on this, which perhaps
is unprofessional (I'm not sure, I'll accept people's opinions
on this in private messages to me) but I must say you took
the bait gloriously......

The provider topology I described is JVNCnet.  Given the state
of the infrastructure in certain places in the Far East and
trans-pacific tariffs and Japanese tariffs and who knows for
what other reasons, they and their Far East customers find it cost
effective to connect to the internet via direct links to the
northeastern USA.  Maybe those are the traces that Dennis
Ferguson put in his message--I don't know for sure.

Now, maybe this is a temporary aberation and as soon as parts of
the far east gets its infrastructure together this topology will
be deemed silly, but I think in general it is dangerous to assume
that the topology will always follow geographical boundaries.....


To be fair, I think that the 64-bit SIP address will work, but
it will require constant management.  I have gone through the
exercise of calculating how many hosts the 64 bit address can
address.  I don't remember the exact numbers, but with five
levels of hierarchy (which I think would be necessary for 10^12
hosts), the SIP address could handle 10^12 hosts if it acheived
about 2% efficiency per level of hierarchy.  By 2% efficiency
I mean that 2% of the possible address assignments (per level)
are actually assigned.

This is really low efficiency, and I think one can do better
if one manages the addresses carefully and allows for
re-assignment of addresses when the previous allocation
strategy proves to have been wrong.  But, without good
allocation it is easy to do worse than 2% (the pre-CIDR
IP address space is probably getting about this kind of
efficiency).

So, I don't think 64-bits is broken, but I think it will
require constant work to keep it useable.  You (Simpson)
claimed to have spent 100 hours on the allocation strategy
you made.  That's a lot of time, and already we have
counter-examples to your strategy.  Address space wars
are really painful, and I think it would be nice if we didn't
have to have them in the future....

PX