IP Switches

[Bdale Garbee <bdale@col.hp.com>]

> While I agree 100% with your comments on the Grace board for systems
> that require 4 or more ports, what about those that REALLY only need
> 2-3 ports?

> Is the DE a viable alternative?

Possibly.  My initial motivation for working on a port of NOS to the Intel
processor standalone environment was motivated by the existence of the PS-186,
which in the current design rev is actually 5 ports... 4 HDLC ports with DMA,
and one "console" port that could talk to an external TNC if so configured.

The Data Engine was interesting for three reasons... it promised to be cheaper
than a PS-186 by about a factor of two, it obviously became available sooner,
and because the V40-based DE design with 8530 is so close from a software
standpoint to the 80C186 with NEC DMA and 8530 design of the PS-186, I
expected to get "two for the price of one", in that the DE would be little
additional effort.  Since that time, the effort Kantronics has put into
developing compatible modem daughter cards and radios for 1200 and 9600 baud,
with 19200 and faster in the future, would appear to be additional motivation,
except that I have only a marginal interest in links that slow.  The big
downer is that because of the way the V40 works, you either get two ports with
full duplex DMA, or one port with full duplex DMA, one with half duplex DMA,
and the "console" port.  While from what you say later that might be a good
match for your needs, it isn't for mine.

Because I had a working DE before I had a working PS-186 prototype, I actually
used the DE for a lot of my development work last spring, but that's irrelevant.


With no word on the PS-186 from AEA in months, I don't know where it stands.
Given that the list price Mike Lamb mentioned to me some months back is about
the same as a Grace standalone card, I don't think I could recommend it in good
conscience to anyone anyway, since the Grace card is much cleaner and faster,
mostly because of the more contemporary processor, viability of DRAM as
demonstrated by the Microsats, etc.  Plus, Don N4PCR and his cohorts have done
all the really hard software work already, and I am, like many hams, basically
very lazy!  :-)  At least to the extent that I'd rather work on things I can't
buy than duplicate effort.

> Are you going to complete NOSINABOX for the DE or are you now putting
> all your effort into the GRACE board?

I am currently focussing all my efforts on the Grace board, primarily because
I am under some peer pressure as the newly formed Colorado Packet Association
aka COPA gets rolling.  I've talked about high speed packet for a long time,
but only have 1200 baud in and out of my basement full time.  It's sort of a
case of "put up or shut up", and the Grace card represents a better match in
terms of number of ports and in terms of the effort I will have to expend to
our needs than anything else at this time.  The initial cost is, I think, very
reasonable given the effort the design and software represent.

I haven't completely given up on the DE, but around here, we need more than
2 ports on the backbone nodes, and we don't have any need for lesser "leaf"
nodes until we have a working backbone, so the DE running NOS just isn't part
of my immediate needs.  If there's enough demand, and especially if there is
any local demand in Colorado, I might find time to get back to it later this
year... but not soon.  If someone else wants to port NOS to the DE before I
do, that would be great!  The tools I was using include Turbo C 2.0, and a
linker/locator called Paradigm LOCATE from Paradigm Systems.  They advertise
in Embedded Systems Programming magazine, and their tools are very nice, though
not cheap.

> Sorry about all the questions :-) but I just spent this last weekend
> convincing NAPRA to completely rebuild two of our sites around
> Data Engines running NOS. These sites only have (and will continue to
> have for many years) three ports.

Keep in mind that a DE is really only two ports, you have to do something
external like a TNC to use the console port as an RF channel.  Kind of ugly.
Either buy Grace cards, resolve to live with G8BPQ code on the DE's, or
port NOS locally would seem to be the options.  The G8BPQ code, by the way, is
available for free from the Kantronics phone-line BBS.  I don't have the number
handy, but it's in lots of their literature.

> While we will have Grace boards at
> several locations, the majority of sites in the Pacific Northwest are
> run by club's or individuals. They only wish to have a single user port
> and a network connection.

It is obvious from this, or relatively obvious at least, that you guys are
still thinking in terms of omnidirectional RF links for your network
connections.  I expect most if not all of ours to be point to point.  Given a
point to point connection mechanism, 3 ports is the practical minimum for a
"tap" in a backbone (2 ports), plus a local access port.  Further, we're going
to try hard to make most if not all of our local access ports full-duplex
repeaters, so that we at least have CSMA between local users.

> We (NAPRA) wish to encourage them to eventually
> run TCP/IP and it seems that the DE may be the lowest cost solution.
> I hope you do continue work on NOSINABOX for the DE, it still has it's
> place, limping along or not.

It is unlikely that I will work on it again soon.  I am not currently motivated
to work on it for free since the Grace card is a much better fit to our local
needs, and it is unclear to me that it is legally or morally possible for me
to charge anything for the port.  That is not a guarantee that I won't work on
it again, just an indication that it's nowhere near the top item on my list
right now.

73 - Bdale, N3EUA