Re: Policy Statement on Address Space Allocations
Scott Huddle writes:
Name a backbone which doesn't come into the San Francisco Bay Area.
Um, all of the rest of the world's providers.
Backbone, Scott. Backbone. You, Sprint, PSI, Alternet, Net-99, etc. All the rest of the world's providers are getting transit from some backbone. If all the transit backbones are in the area the problem is merely political.
They aren't... ICM, Pipex, and Dante to name three. Sprintlink may play nice and handle ICM, what do you plan to do to address the others? And for the next trick, how do you scale your solution to the next site? Does your solution require sites of all the backbone providers to be at each metropolitan exchange? Doesn't this put a limit in the number of providers that can do this? -scott
Backbone, Scott. Backbone. You, Sprint, PSI, Alternet, Net-99, etc. All the rest of the world's providers are getting transit from some backbone. If all the transit backbones are in the area the problem is merely political.
They aren't... ICM, Pipex, and Dante to name three. Sprintlink may play nice and handle ICM, what do you plan to do to address the others?
How do Pipex and Dante get global routes right now?
And for the next trick, how do you scale your solution to the next site? Does your solution require sites of all the backbone providers to be at each metropolitan exchange? Doesn't this put a limit in the number of providers that can do this?
More likely, it limits the number of areas you can apply this idea cleanly to. Poorly-connected areas won't get such blocks. The more backbones in an area, the easier (technically and politically) to put such a block there. The question is how much you get out of the areas we can do this to... which could be quite a lot. Just the SF Bay Area is a large chunk of the Internet as a whole... it won't be forever, but it is now and its growth patterns could positively or negatively shape how other areas grow later. -george
In message <199602012310.AA15971@mail.crl.com>, George Herbert writes:
Backbone, Scott. Backbone. You, Sprint, PSI, Alternet, Net-99, etc. All the rest of the world's providers are getting transit from some backbone. If all the transit backbones are in the area the problem is merely political.
They aren't... ICM, Pipex, and Dante to name three. Sprintlink may play nice and handle ICM, what do you plan to do to address the others?
How do Pipex and Dante get global routes right now?
Oh.. from a whole bunch of different places. ANS customer routes from ANS MCI customer routes from MCI SprintLink customer routes from Sprint Alternet customer routes from Alternet AUNET routes from EUNET ... You get the idea. And each of these I'm sure has routes all over the 192/8 space. Curtis
How do Pipex and Dante get global routes right now?
Oh.. from a whole bunch of different places.
ANS customer routes from ANS MCI customer routes from MCI SprintLink customer routes from Sprint Alternet customer routes from Alternet AUNET routes from EUNET ...
You get the idea.
How do they get to Barrnet? SURAnet? Etc? There are already geographically limited large providers, they just aren't well aggregated. The transit problems have been dealt with before and will be dealt with again. -george
How do Pipex and Dante get global routes right now? Oh.. from a whole bunch of different places.
ANS customer routes from ANS MCI customer routes from MCI SprintLink customer routes from Sprint Alternet customer routes from Alternet AUNET routes from EUNET ...
You get the idea.
How do they get to Barrnet? SURAnet? Etc?
I'd bet the bbnplanet peer at mae-east, if i had a clue. Its the closest touchdown where they all meet. --bill
In message <199602021946.AA16364@mail.crl.com>, George Herbert writes:
How do Pipex and Dante get global routes right now?
Oh.. from a whole bunch of different places.
ANS customer routes from ANS MCI customer routes from MCI SprintLink customer routes from Sprint Alternet customer routes from Alternet AUNET routes from EUNET ...
You get the idea.
How do they get to Barrnet? SURAnet? Etc?
Suranet is at MeaEast. Barrnet is presently an MCI transit customer.
There are already geographically limited large providers, they just aren't well aggregated. The transit problems have been dealt with before and will be dealt with again.
Yes and if every geographically limited provider contracted the same transit provider your proposal would have no holes in it, but they didn't.
-george
Curtis
Suranet is at MeaEast. Barrnet is presently an MCI transit customer.
There are already geographically limited large providers, they just aren't well aggregated. The transit problems have been dealt with before and will be dealt with again.
Yes and if every geographically limited provider contracted the same transit provider your proposal would have no holes in it, but they didn't.
That doesn't matter. What matters is if the geographically limited providers contracted for transit with providers that touch down in the area or not. If yes, no problem. If no, then we have a dead zone and need to figure out how to get the routing working. MCI is in the Bay Area. I don't know what we'd do for Sura; are they paying someone for generalized transit, or just peering? -george william herbert gherbert@crl.com
In message <199602030127.AA03799@mail.crl.com>, George Herbert writes:
MCI is in the Bay Area. I don't know what we'd do for Sura; are they paying someone for generalized transit, or just peering?
You question was how Pipex reaches Suranet. Suranet reaches the west coast via MCI. The transit issue is who transits the portion of 192/8 to providers like Pipex that just touch the East coast and don't have transit to the West coast, just access to direct customers of US providers who are at MaeEast (and for some MaeEast and Sprint NAP). A peering arrangement at a US West coast interconnect and no US transit agreement does not get you transit to a lot of places. If all of the major providers send traffic to 192/8, someone then has to provide transit to the portion of 192/8 that you can't get to from the West coast without a transit arrangement. Alternately, you could just require all of the providers who peer with someone with no transit agreement to the West coast to carry all the 192/8 prefixes for that peer only. What you then have is an agreement to do cross provider aggregation with any exchange interprovider traffic for a specific aggregate at a specific point only, with consideration given to the holes in the aggregate by every participating provider. This can be done in principle but would require more control over configuration than some providers are currently capable of. If you further scale back the idea to only providers whose router configuration is sufficiently automated to handle this, you start with the null set today, with one or two providers talking in the IETF RPS WG about being able to do this in a short while (which seems to be working out to be a large value of short:).
-george william herbert gherbert@crl.com
Curtis
participants (4)
-
bmanning@ISI.EDU
-
Curtis Villamizar
-
George Herbert
-
Scott Huddle