I-D Action:draft-azinger-scalable-addressing-00.txt

[Brian E Carpenter]

I think the message below captures the problem with the multi-PA
model perfectly. When we started out in this direction almost
15 years ago, exit router selection didn't seem like a big deal,
but now it does. Solutions such as ILNP and LISP do offer a way out
of this, but for short term deployment decisions, they're irrelevant.

Regards
   Brian Carpenter

On 2010-09-29 20:41, S.P.Zeidler wrote:
> Thus wrote Brian E Carpenter (brian.e.carpenter at gmail.com):
> 
>> The problem is, site IT managers don't seem to like this model and its
>> implication of renumbering when you add or drop an ISP.
> 
> I'm currently a network admin 2 days a month, accumulated, the rest of the
> time I'm a sysadmin. The thing -I- seriously don't like about this model
> is that it puts the decision about routing on the communication endpoints,
> all of them, in their precious multitude and variability.
> 
> The model expects that every last desktop OS has a perfect and
> sophisticated network stack, and that it will remain bugless forever
> and anon even in rarely used variants. This is quite unrealistic.
> As a result, you'd turn network admin into a job that scales with the
> number of hosts, not with the number of routers.
> 
> Balance the cost of this manpower against the cost of running BGP.
> With enough hosts, BGP will be cheaper -> all sites above a certain size
> that require resilient networking will want their own presence in the DFZ.
> I'd guesstimate the flip point at about ~100 hosts that actually need to
> have connectivity (as opposed to, need to be able to talk to a web proxy
> and mail server).
> 
> regards,
> 	spz