[ppml] Provider Independence???

[marcelo bagnulo braun]

I think we are mixing stuff here, that are slightly different.
I think that we have different approaches here:
one thing is geo addressing as in Tony Hain's PI proposal. In this 
case, the address block you get depends on your GPS location
another different thing is IX based addressing, that can be considered 
a form of geo addressing but with additional constraints. In this case, 
the address block is assigned to the IX, and a site can use it as long 
as its ISP is in the IX. If the site changes to another provider that 
is not in the IX, it needs to change addresses (even though the site 
haven't moved geographically)

I thought we were talking about the first one, but now it seems we are 
considering the second one. Could you confirm what are we talking about 
here?

In IX based addressing there is no problem with topology matching 
geography since this is the case by definition. At this point, the 
sites would obtain an address block directly from the IX and the ISP of 
the IX that want to provide service to one of such users need to 
advertise the whole IX address block. So all isps announcing the 
address block will receive traffic to any of the sites that are using 
IX addresses. It is up to the ISPs to decide whether they want to serve 
such sites or not. I still think that this approach doesn't make 
business sense. But i think that we repeating ourselves by now, so i 
will shut up now.

regards, marcelo



El 09/12/2004, a las 18:32, Michael.Dillon at radianz.com escribió:

>>>> but what happens if a ISP don't want to accept the peering 
>>>> conditions
>>>> of the other ISPs in the region, what happens then?
>>>
>>> Then they don't use geo addresses.
>>
>> i am really not following...
>
> Geo addresses are just ordinary IPv6 addresses that
> work the same as any other IPv6 addresses. They differ
> from the current IPv6 addresses in only two ways. One
> is that they come from a well-known prefix so that
> anyone who chooses to can apply different policies
> to these addresses. The other difference is that the
> RIRs allocate these addresses according to a geographical
> hierarchy and pass on the requirement to XPs to maintain
> this geographical hierarchy.
>
> This means that the allocation to an endpoint does not
> come from their provider. It is provider independent.
> Instead, the allocation comes from the local exchange
> point operator. At this point, these addresses are
> usable through any provider connected to the XP
> and the routing announcements for these addresses
> can be aggregated globally to a very small number
> of entries in the so-called global routing table.
>
> This doesn't prevent anyone from continuing to
> use provider allocated IPv6 addresses in the same
> way that they are used in the v4 Internet.
>
> The ITU can make credible arguments that the
> existing random allocation of IPv4 addresses
> will not scale in IPv6 and therefore an alternative
> is needed. If the Internet community does not
> provide some type of geographic addressing that
> as an alternative to the ITU's proposal, then
> the ITU's position can easily win by default.
> We can no longer ignore the phone network because
> in the 21st century, the Internet *IS* the phone
> network. I disagree with the ITU position.
>
> --Michael Dillon
>