Frank,
> Shiv,
>
> The requirements document specifies only that any new routing
> and addressing architecture/system support the hop-by-hop
> model currently in use. It is intentionally silent on other
> models. It does not require other models be used nor does it
> prohibit them from being used.
>
> That said, a personal observation is that the probability of
> a new architecture/system being deployed is inversely proportional
> to the amount of any required changes to end-systems. In fact, the
> probability may well be a step function -- if you have 0.000 changes
> the probability of successful deployment is 10%, if you have >0.000
> required changes, the probability of successful deployment goes to
> 0.0000% :-(
>
> Similarly, if a proposed new architecture/system _requires_ changes
> to the forwarding path of routers, the probability of successful
> deployment approaches 0.00 because of the widespread use of ASICs
> in high-speed forwarding paths. ASICs have a rather long and
> expensive edit-compile-debug loop...
While the above is true in principle, let's look at the practice.
In addition to the presently used hop-by-hop forwarding paradigm,
the only other unicast forwarding paradigm known today is the
"explicit" routing. Given that explicit routing is supported *today*
(via MPLS) *at the line rate* by quite a few router vendors,
requiring support for explicit routing isn't going to affect the
probability of successful deployment of the new architecture.
Yakov.
> But those are just personal observations.
>
> At 06:59 PM 12/18/01 -0500, Shivkumar Kalyanaraman wrote:
>
> >FOlks,
> >
> >The requirements doc does not say anything about whether changing in the
> >routing system for the future should or should not entail changes in the
> >forwarding paradigm.
> >
> >History has shown that CIDR, subnet masking and MPLS-style
> >significant control plane changes have involved changes in the data plane,
> >and in some cases mandatory...
> >
> >I believe we should leave the door open for such changes, but carefully
> >police them
> >
> >
> >-Shiv
> >===
> >Shivkumar Kalyanaraman
> >Associate Professor, Dept of ECSE, Rensselaer Polytechnic Institute (RPI)
>
This archive was generated by hypermail 2b29 : Mon Aug 04 2003 - 04:10:03 EDT