roy@phri.UUCP (Roy Smith) (12/07/87)
A couple of weeks ago, I asked what people thought about my idea of
rolling my own bargain-basement etherbridge by putting two Kinetics KFPS-2's
back-to-back on opposite ends of a LADC (4-wire leased) circuit, with Farallon
PhoneNet drivers. Thanks to the following people for responding:
husc6!sob (Scott Bradner)
ucscc.UCSC.EDU!haynes (Jim Haynes)
bu-it.bu.edu!ccjap (John Papadopoulos)
bu-cs!bu-it.bu.edu!kwe (???)
reed!goetz (Norman Goetz)
I said that the PhoneNet could drive up to 4000 feet of twisted-pair
cable, and that I hoped that our 2000-foot line-of-sight circuit would be
withing that limit as-the-cable-snakes. Several people pointed out that phone
lines run radially from the CO, with an LADC circuit consisting of two pairs
(one from here to the CO, the other from there to the CO) patched together,
and thus my 4000-foot hope was unrealistic. Perhaps, but I suspect that the
people who suggested that have never been to Manhattan. CO's are every few
blocks around here. I'm still hoping I'll make the 4000-foot limit, but I'm
not holding my breath.
Anyway, with much random editing, some of the more interesting
comments.
----------------
From: allegra!likewise!uunet!ucscc.UCSC.EDU!haynes (Jim Haynes)
Well, a guy here has been trying it for months and it still hasn't worked. I
guess he doesn't have 'appropriate software' to download. The problem, as I
understand it, is that the Kinetics software isn't prepared to cope with
having two Kinetics boxes on the same Appletalk line.
----------------
From: cmcl2!harvard!husc6!bu-cs!bu-it.bu.edu!kwe
It's crazy enough that it just might work! It's a novel idea.
I like your inventiveness.
I'm not sure you want to risk the investment on the Kboxes,
since I would guess that the Appletalk link will not work. If you
have the need for the Kboxes anyway, I would recommend trying the
set-up you suggest.
You can run anything you want on a LADC, but the telco will
only guarantee (and it's a weak guarantee) service up to 9600 baud.
19.2k is usually worth a try, 38k is very iffy, and 56k is not going to
cut it. 230k is way out there.
You can get a good idea of how long your circuit is by knowing
where your Central Office is physically located. If it's two miles
away, you have a four mile LADC (20k ft). All local circuits
originate in a CO, that's the definition of a CO. There is no
patching on the pole. Chances are slim you will get a 4k ft circuit.
But I like the idea anyway.
----------------
From: cmcl2!rutgers!mimsy.umd.edu!uunet!tektronix!reed!goetz (Norman Goetz)
I think the Farralon hardware would work fine this way. [...]
Unfortunately I don't think this will work. The hardware connections
are fine but the Kinetics boxes are neither routers nor half-repeaters
which is the role you are asking them to play. I've heard it described
that the K-boxes are not true IP gateways but more remote front ends
for devices on AppleTalk.
Have you looked into twisted-pair Ethernet as a link?
----------------
In addition, I had a conversation with Bill Russell (russell@nyu.edu)
who pointed out that even if the LADC could support AppleTalk over whatever
length the circuit turns out to be, and if the boxes could be taught to be
proper level-2 IP bridges, there is the problem of speed. We were talking
about Ungermann-Bass etherbridges, which would cost about $16k for a pair,
running 56 kbps over the LADC. Why so much, I asked, when it's obvious you
can put together a reasonable stab at the same end result for about a quarter
the price, and at five times the serial line speed? Bill's answer was that
the limiting factor was not line speed, but CPU cycles. The U-B bridge can
pass hundreds of packets per second. A kbox, with a poor little 6809 (?)
processor in it, couldn't hope to keep up with a fraction of that traffic, not
to mention the minimal amount of ram available (i.e. minimal routing tables,
minimal packet buffering, etc).
--
Roy Smith, {allegra,cmcl2,philabs}!phri!roy
System Administrator, Public Health Research Institute
455 First Avenue, New York, NY 10016roy@phri.UUCP (Roy Smith) (12/08/87)
In article <3057@phri.UUCP> roy@phri.UUCP I wrote: > The U-B bridge can pass hundreds of packets per second. A kbox, with a > poor little 6809 (?) processor in it, couldn't hope to keep up with a > fraction of that traffic. Two corrections to add. First, I got a note from kinetics!swan (Michael J. Swan), who is Director of Software Products. Michael says: [...] no, that type of setup wouldn't work because our gateway is not a true IP router (I assume you were routing IP style packets). The KFPS code routes according to what's commonly refered to as Mac/IP routing which implements a subset of IP-style routing. Also, to dispell some other misinformation you received, the KFPS uses a 68008 processor, not a 6809. Also, from glancing at my newly-arrived Ungermann-Bass product literature, I see that my estimate of "hundreds of packets per second" was low by about an order of magnitude. -- Roy Smith, {allegra,cmcl2,philabs}!phri!roy System Administrator, Public Health Research Institute 455 First Avenue, New York, NY 10016
berner@apple.UUCP (William Berner) (12/12/87)
Roy smith posted a couple of message about using 2 Kinetics boxxes as an Ethernet bridge. However, if you're just looking for an Ethernet bridge, then Retix sells one called the RetixGate 2255, that's priced at less than $2000. You can find out more by calling Retix at 800/255-2333 Bill Berner Apple Computer, Inc. Desktop Communications Marketing