Re: Experimental RFC to be: draft-stoica-diffserv-dps-01.txt

[Ion Stoica <istoica@cs.berkeley.edu>]

Bill,

I do understand that configuration is a big problem in practice,
but I don't see how implementing DPS otherwise -- for example,
by adding a new label between IP header and the link layer/MPLS header
-- would eliminate the problem. If the egress router "forgets" to remove
the label,
the end-host would be equally confused. And isn't this the same problem
that MPLS has?

Ion

Bill Fenner wrote:

> >This can happen only if the eggres node of a DPS domain
> >"forgets" to reset the ip_off value to zero and update the IP checksum.
> >But this is a very simple operation, and I don't see many ways one
> >can get it wrong!
>
> Ion,
>
>   One of the big problems in router software architecture is
> configuration and its complexity.  Mark Handley made a good analogy --
> today's router configurations are mostly like programming a processor in
> hex.  You have to get everything right, and you have to consider
> everything individually.  You don't even get functions or macros.
> Therefore, consider that routers probably won't be configured "This
> router is a DPS domain border", rather perhaps "perform the DPS domain
> border functionality on packets of this type" or even "Zero out the
> ip_off field on these packets".  It's easy to imagine some important
> bit of the configuration being forgotten.
>
>   Bill