Cable Modem [really more about PPPoE]

Christian Kuhtz ck at arch.bellsouth.net
Tue Jun 26 17:02:53 UTC 2001


On Tue, Jun 26, 2001 at 12:40:54PM -0400, Daniel Senie wrote:
[..]
> This isn't necessarily the case. Take a look at rent-a-POP dialups. UUNet 
> sells services to LOTS of ISPs from all of its POPs. The RADIUS 
> authentication has to get to the proper ISP, but the traffic certainly does 
> NOT travel to the ISP's network before going out into the world. 
> Traceroutes from your dialup account will show that the traffic goes 
> directly. UUNet charges the ISP a fee for the use of the dialup, and UUNet 
> provides a relaying of the RADIUS traffic to the right ISP, and provides 
> transport for the end user's packets.

.. which btw could be anything.  Could be proxy RADIUS or proxy DHCP or
whatever.
 
> The same scenario COULD be used in DSL or cable setups. That's not to say 
> that it will be.

If I may add my $.02...

I think you hit the nail right on the head.  And cost pressure is what *may* 
change that COULD to a WILL. (Seems things have to get a lot worse before they
will get really better).

Seems that the model works for dial because nobody has any ill conceived 
notions about needing to 'own' the last mile like connection to the subscriber
to offer value add services.  For two reasons: a) people have realized that 
there really isn't much value in owning dial POPs anymore and b) it's difficult
to add much of any value add services to the link itself because of the limited
bandwidth. 

This notion of selling value add services and 'owning' the subscriber is 
something very commonly found in broadband networks.  

Sure, you could send broadcast quality TV down a DSL pipe, but who's actually
doing this as a generally available service? 

Dial appears to be such a low end commodity that people have given up on that 
type of model, and just want to get folks on the net with a certain quality of
access but that's about it.  Isn't that the one thing what most if not all
DSL and cable lines achieve?

Seems to me that there's too much marketecture driving these deployments, and
that the rent-a-POP model could drive cost down considerably.  Which would mean
that all of us would be better off for it.  Those who object to rent-a-POP 
models typically are the same that think # of hops matters vs actual 
latency/quality of service/access and believe that video etc services must
always be supplied locally to be working properly.  Again, clueless 
marketecture methinks.

I don't claim to have a solution, I just find it interesting that there appears
to be a fairly narrow horizon in this area, when it has clearly been proven
that folks are willing to live with the other model just fine.  The rent-a-POP
model's sole weakness is an organization's inability to work out a proper
business model...

Now, what does all this have to do with DHCP or RADIUS?  Well, in the 
rent-a-POP model, nobody cares how the actual authentication and accounting is
done as long as it is accurate and fits the billing system needs of the 
wholesale customer.  Whatever works, is relative to the other options cheapest
and most reliable (offering the greatest economies of scale) is the one that's
chosen...

Do I carry a religion either way?  Not really.  Seems that DHCP & DynDNS 
integration is more readily available than RADIUS & DynDNS... But that's just
my opinion and not a representative survey.

Cheers,
Chris

-- 
Christian Kuhtz <ck at arch.bellsouth.net> -wk, <ck at gnu.org> -hm
Sr. Architect, Engineering & Architecture, BellSouth.net, Atlanta, GA, U.S.
"I speak for myself only.""



More information about the NANOG mailing list