On 2009-01-22 10:32, David Meyer wrote:
>>> Brian, you know very well that we can test and evaluate new technology
>>> in various ways before its deployed.
>> Of course; interop testing (not carried out officially by the IETF) is
>> standard operating procedure, although not formally required until
>> you get to DS level. But my point is that the real question about
>> LISP is not whether it will work at small scale, but whether it will
>> work reliably and efficiently on an Internet with millions of active
>> prefixes (as opposed to 300k prefixes).
>
> Brian,
>
> How would you envision such a result being "shown"? Here
> I'm asking not specifically about LISP (although LISP
> would seem to make a fine test case), but more generally
> what methodology will yield a relevant result here (i.e.,
> answer the question you pose)?
That's quite a profound question, I think. Apart from coming
back in hundred years to see if BGP4 is still in use, there's
surely no simple answer. None of the literature I've read on
studies of the evolution of the routing system really gives me
any sense of a predictive model for the topology. Without a
predictive model, it's very hard to analyse scaling over a few
orders of magnitude. And even if we had a model, deployment of
a loc/id split would change the model in a non-linear way.
Again, more of an RRG topic, I guess.
Brian
_______________________________________________
Int-area mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/int-area
