The IP over ATM Mailing List Archive by date

Cell Relay Retreat>List Archive>month:1995-Apr> msg00050



[Date Prev][Date Next][Thread Prev][Thread Next]  
  [Date Index][Thread Index][Author Index][Subject Index]

Flame re: "routers should be a dying breed"

  • From: Masataka Ohta <mohta@necom830.cc.titech.ac.jp>
  • Date: Tue, 25 Apr 95 14:08:07 JST
  • CC: ghill@atmsys.com, laubach@terra.com21.com, ip-atm@matmos.hpl.hp.com

>     > Personally, our architecture should permit hosts and routers to take equa
>    l
>     > unbiased advantage of "mesh" or "server" mode. 
>     > 
> 
>     I don't agree here - in ATM land, routers should be a dying breed (just for
>     legacy/ethernet stuff), and we should optimize for directly connected hosts
>    .
> 
> <Ignite Rear Afterburners....>
> 
> Asserting that routers are a dying breed is wishful thinking, not technical
> analysis.  At this point, it is clear that high-speed networking technologies
> other than ATM (e.g., 100 Mb/s Ethernet, and at least one of FibreChannel
> and HIPPI) will have a place in future networks.

Why do you discriminate ATM?

Surely, legacy ATM switches can't be IP routers.

But it is a problem of the legacy ATM and is already solved. See minutes
of colip ATM BOF at Danvers IETF.

> And the right way to
> connect diverse networks is with routers.  If we don't design for routers
> now, we'll find ourselves having to put them back in later, at higher cost.

Not later but now, ATM peopl are just designing a router equivalent and
call it a switch. The problem is that such switches use ITU-derived
NSAP based protocols, which can not directly interoperate with the
Internet.

Not surprisingly, ATM switches, or anything which is as complex
as IP routers,  need at least as much configuration effort as
IP routers.

Moreover, on the multi-protocl Internet, configurration for both
legacy ATM switches and IP routers is a lot more complex than twice
that of IP-router-only enviroment.

With legacy ATM architecture, non-router hosts also needs extra
configuration.

> I also believe that the widespread and indiscriminate use of the term "legacy" 
> to dismissively describe any technology (however modern) that doesn't fit
> the limited ATM mindset is a great disservice to the networking community.

"legacy", here, means "architected toward legacy phone networks and not
friendly to the Internet", of course.

							Masataka Ohta