NetNews Usenet Archive 1993 #3/NN_1993_3.iso/spool/comp/dcom/isdn/1212

Internet Message Format | 1993-01-23 | 3.7 KB

Path: sparky!uunet!olivea!sgigate!sgi!rhyolite!vjs
From: [e-mail] (Vernon Schryver)
Newsgroups: comp.dcom.isdn
Subject: Re: Low cost ether/isdn brouters (was PC-NFS PPP Serial/ISDN driver wanted)
Message-ID: <[e-mail]>
Date: 22 Jan 93 20:05:41 GMT
References: <[e-mail]> <[e-mail]>
Organization: Silicon Graphics, Inc.  Mountain View, CA
Lines: 74


In article <[e-mail]>, [e-mail] (Tom Dietrich) writes:
> ...
> On low speed links (64Kbps and below) bridging is going to be a
> problem from network to network. The background traffic (broadcast and
> multicast) tend to take up enough bandwidth to make "real" traffic too
> slow to use....

> I've spent a lot of time working with customers who try to save the
> cost and complexity of a router by installing a bridge. Most have
> regretted it.

That is an interesting (to me, of the router faith) datum.


> [e-mail] (Wolfgang S. Rupprecht) writes:
>...
> >Not to start a war or anything, but I'm curious why do folks build
> >bridges as opposed to routers.  Is it to save the cost of writing the
> >code to do the TCP/IP hacking or is it that they would need much more
> >hardware to support the computional horsepower needed to unravel the
> >upper level protocol and then wrap the data-packet back up?


> The cost of producing a router is greater than a bridge. There are a
> lot more decisions to be made in a routing situation, and multiple
> protocols make it worse. Bridges are simpler to make, install and
> administer. There are tradeoffs.

One justification for bridging are stupid, ill-designed protocols that
cannot be routed.  LAT is an infamous example.

Still, the distinction between "router" and "bridge" is exaggerated,
and usually most exaggerated by those of the bridge faith.  A "bridge"
is nothing more or less than a router that can only route "MAC
packets".  Modern bridges even have a "routing protocol," although they
call it "spanning tree."

Routers do not need entire protocol stacks.  For example, to route IP a
router needs to know only about IP, a very simple protocol, nearly as
simple as ethernet.  An IP router need know nothing about TCP, UDP,
telnet, or any other higher layer protocol.  On the other hand, modern,
full-feature "bridges" have complete TCP/IP protocol stacks so that
they can be "managed via SNMP."

Some people bogusly claim that bridges can do things that routers
cannot.  A typical example is "cut through routing", forwarding a
packet before it has been entirely received.  That is equally possible
in a bridge and a router.  A bridge is only a "MAC router".

Many, perhaps most of the bridge faith have an excessive fear of
routing, and often go on about vast quantities of code and the
difficulties in finding it, entirely ignoring things like ka9q.  This
fear probably comes from the fact that simple bridges were often
primarily designed and built by hardware designers, and those guys have
historically had a lot of fear and loathing of software.  What with
verilog, PAL's, "silicon compilers", and other modern hardware design
stuff, hardware guys are gradually being forced to lose their soft
phobias.


Back to the original thread, most of us don't want to buy an ISDN
bridge.  I think few of us want to forward Appletalk or any protocol
other than IP over our ISDN links.  Most of us want a simple, cheap IP
router, probably one that does compression to make 64Kb seem less
slow.

If the established (hardware) companies don't build one, then history
will repeat itself.  Remember that Telebit and cisco started (I think)
as largely software efforts.  Now "brouters" are common and all modems
have lots of software.


Vernon Schryver,  [e-mail]