a radical proposal (Re: protocols that don't meet the need...)
Paul Jakma
paul at clubi.ie
Fri Feb 17 02:05:23 UTC 2006
On Thu, 16 Feb 2006, Vince Fuller wrote:
> to two popular "geo-topo" addressing domains, say the Bay Area and
> the DC area. Let's say that 10.0.0.0/8 is the "geo-topo" address
> block in the Bay Area and 172.16.0.0/12 is the "geo-topo" block in
> the DC area. This provider has four customers in the Bay Area:
>
> 10.1.1.0/24
> 10.10.4.0/22
> 10.100.8.0/21
> 10.200.0.0/16
> customers. For him to provide connectivity to all the address
> range, he must
> a) have full routing connectivity to all other providers that have
> addresses in the same range; this implies that he connects to all IXs
> within the region and maintaines a full-mesh of routing information
> (today, BGP sessions) to all of these providers
That's not quite correct. They would have to:
a) Have full routing connectivity to all other providers who
provide transit in/out of the area concerned.
It does not imply:
- having to peer with every provider in the area (some
providers may be wholly within the area, you wouldn't need
to peer with them, only their 'transit provider')
- having to peer at every IX (you only need to fulfill
condition a)
- that peering with the other providers who provide
inter-geo-area service, with whom you must peer as per a,
must occur locally - it does not. (e.g. you could hand-off
ACME providers Bay Area prefixes to ACME at DC if you
want).
> b) must be willing to provide connectivity to all sites within the region
> to any place that he advertises the prefix 10.0.0.0/8
Right.
> through routing
> exchanges; if he advertises this prefix to non-customers, it implies
> that he is will provide free transit to his competitors' customers
> which are numbered out of this block
That's not correct. Nothing says it has to be free.
If you're handing off X GiB of 10/8 Bay Area traffic to ACME provider
each day, then you would (presumably) charge ACME your costs for
those X GiB. ACME presumably would do likewise for traffic to 10/8
they carried that happened to be one of your customers instead.
So it's normal peering business; indeed it could be a beneficial
business model to try carry as much of that 10/8 traffic as possible.
Some upsides:
- scenic routing would be far less prevalent.
- trivial provider-changing for customers / much increased
competition (easier to attract new customers away from other
providers).
Some big downsides:
- trivial provider-changing for customers (your competitors can
get your customers to change over more easily than today) (I
suspect providers would be more wary of this than they would
welcome the /increase/ in competition ;) ).
- every customer's (using these geo-assigned addresses) traffic is
dependent on every transit provider. So ACMEs' customer could face
an outage because "Barr's Internet Services" has a failure. This
could be mitigated with good practices (ensure that those providers
who provide transit into the area only ever originate the
area-prefix from within the area, never outside - hard to know how
that could be enforced)
- Co-ordination of origination the prefix: How do you ensure that
those providers who announce the 10/8 prefix are only those
providers who are peered with all the others? Squabbles could get
really ugly and affect /all/ users in that block, regardless of
whether they are customers of the squabbling providers.
> "Addressing can follow topology or topology can follow addressing.
> Choose one."
>
> and I'd offer a corollary:
>
> Transit relationships (i.e money) must follow topological relationships
> (and thus addressing); the alternative is some combination of inefficient
> or non-scalable routing, black holes, settlements, regulation, or other
> undesireable things.
We have settlements today already. The money factor isn't a problem
really - seems to me at least the money aspect could work fine for
geo-addressing, as it (should) do for transit services today. It's
the other inter-provider co-ordination problems that would make it
problematic.
There'd need be someone who could "enforce the law", after defining
the "law" of course ;). Though, we happen to have such a body in my
country funnily enough.
> If you really want to combine transport identifier and routing
> locator into a single "address", you give up a lot of flexibility.
> For routing to scale, addressing must follow topology, so in such a
> network architecture the term "topology independent address" (aka
> "provider independent address") is truly an oxymoron.
Right.
The logical step then is for leaf-sites to build upon this
topology-addressed network and advertise the lists of "topology
identifiers" by which they are reachable to each other: shim6. Smart
hosts communicating over a dumb network.
Providers aren't happy with that either though, judging by some of
the grumbling wrt shim6. But that's the only solution left unless
some new 'break-through' solution is discovered.
regards,
--
Paul Jakma paul at clubi.ie paul at jakma.org Key ID: 64A2FF6A
Fortune:
Gold's Law:
If the shoe fits, it's ugly.
More information about the NANOG
mailing list