Human-Nets-Request%rutgers@brl-bmd.UUCP (Human-Nets-Request@rutgers) (01/08/84)
HUMAN-NETS Digest Sunday, 8 Jan 1984 Volume 7 : Issue 5
Today's Topics:
Computer Networks - Networks, Networks, Everywhere (5 msgs),
Input Devices - Dvorak Keyboards &
Keypads (2 msgs)
----------------------------------------------------------------------
Date: 06-Jan-1984 1256
From: John Covert <decwrl!rhea!castor!covert@Shasta>
Subject: A bit more info on Digital's ENET
First I'd like to thank the author of the compendium on networks.
And second, I'd like to give a little more information on the Digital
ENET. It is composed of systems running our DECNET software products,
first introduced about nine years ago.
DECNET is much more than a mail network. It is a product built on a
layered network architecture (DNA) with lower, non-programmer
accessible data-link and routing layers, and higher, programmer
accessible, session layers.
It is similar to the ISO model on open systems interconnect. Since it
is older than that model, it does not correspond exactly, but will,
more and more, as time goes by and as the worldwide networks develop.
At the data-link level it can use synchronous or asynchronous lines of
any speed running DDCMP, public network lines running X.25, parallel
links running protocols specific to those devices, and Ethernet.
Using gateway products it can create gateway links into an IBM SNA
network.
At the user accessible layer, it is possible for any program to open a
transparent, full-duplex, channel to any other program on the same or
any other node in the network. Programmers can take advantage of this
"network logical link" to build any application they wish.
Various Digital supported protocols running on logical links are
host-to-host terminal connections, allowing a user at any node to act
as an interactive terminal on any other node, Mail, the Data Access
Protocol, (see next paragraph) and several others.
The DAP protocol is used to copy files, but it is much more than a
file copy protocol. It permits a program on any system to access a
file on any other system as though that file were a local file. In
fact, VMS and RSX using the DAP routines buried in RMS permit a
nodename to be simply a part of a file spec used by any program.
DECNET does a bit more than implicit routing; it does dynamic path
routing. As a result, given sufficient alternate paths, the loss of
an intermediate node does not affect the operation of traffic
currently routing through that node. Dynamic path routing was first
made available in DECNET Phase III, offered for sale almost five years
ago.
For example, since our network has three transatlantic links, a few
months ago, we had a serious failure of the links between
Massachusetts and the remainder of our engineering and marketing
headquarters 30 miles to the north in New Hampshire. But due to the
fact that some of our transatlantic links go into New Hampshire and
others into Maynard, we did not immediately notice the problem.
Things got a bit slower, since we were no longer using several 56Kbps
links but were pushing all traffic through some 9600bps links to the
U.K., down to Geneva, and back.
The reason there occasionally appears to be some implicit routing in
our node strings is that the Phase III version of DECNET had a maximum
of 256 addresses. This restriction has been lifted in Phase IV.
However, as a result of the restriction, it was necessary for us to
partition our network.
Reassigning node numbers will not be complete for several months, and
not all systems will upgrade, so there may be a few systems which
require one intermediate hop from RHEA. Many of these will have
definitions on RHEA making that transparent to the sender (though a
recipient would see the hop). The rest should be directly addressable
from RHEA, whether located in the U.S., Canada, the Caribbean, Europe
(13 countries now), the Middle East, the Far East, or Australia.
(Remember, IBM is the only computer manufacturer larger than Digital.)
/john
------------------------------
Date: Friday, 6 Jan 1984 10:34-PST
Subject: More on CSNET
From: obrien@rand-unix
Compliments to Mr. Fair - an excellent summary article. Would
that Human-Nets had more such.
To expand on CSNET: It is currently funded by the NSF, and
expects to become self-supporting during the next few years, based on
member fees. These fees are:
$ 30,000 - commercial sites
10,000 - government and not-for-profit
5,000 - educational
These fees may be reduced by petitioning for a reduction in
the case of small outfits, and are lower for people who already have a
net connection via Arpanet.
The CSNET membership list as of Dec. 1 shows:
85 Phonenet sites
6 Telenet sites
18 Arpanet sites
4 CSNET-owned hosts
Not all of these sites are operational yet, though most are.
Phonenet sites are served by two Relay machines, which call them up
nightly to exchange mail. Text files may be automatically transferred
using MMDF-based mail-receipt programs, though this is obviously not
the best way to do business. Bandwidth here is limited by the
1200-baud phone lines as well as by the capacities of the Relays.
Mailing-list stuff can be handled OK, but Usenet traffic breaks the
Relays by sheer load.
Telenet sites run TCP/IP on top of X.25 virtual circuits,
using software developed for CSNET at Purdue. Personally I think this
is hot stuff. If your phone bills are $1500/month, you can run
equivalent traffic over Telenet for about $1200/month, last time we
figured it out. And, you get full Internet connectivity and services
into the bargain. Because the drop lines from Telenet to the host are
really only 9600, 4800, or 1200 baud dedicated phone lines,
instantaneous bandwidth is not as good as Arpanet, but it's not bad.
And, you and the rest of the world will be hard-put to tell that
you're not on Arpanet directly, except you don't have to deal with the
DoD. This software really works, and works well.
Arpanet sites run standard Arpanet software - no change.
In addition to simple net connectivity, CSNET brings the
benefits of centralized network management. Basically this means that
if your mail isn't moving, you have experts to scream to, and they
really will work hard to fix the problem. There are other benefits
such as ongoing mail system development, an automatic nameserver, and
so forth.
Management of CSNET has recently been transferred away from
the contractor committees which built the net to a newly-formed
Executive Committee, which is overseeing the move from a research to a
service organization. The two relay machines are moving to BBN - it's
cheaper and easier to run a single computer center and communicate via
WATS lines than to spread out the Relay operations.
------------------------------
Date: 6 Jan 1984 1808-EST
From: John R. Covert <RSX-DEV at DEC-MARLBORO>
Subject: DECNET and ENET
Just to clarify something... DECNET is the name of a product sold by
Digital which any customer can use to build their own network.
DECNET is used to build Digital's internal network. The internal
network name has been a hotly debated subject (what's in a name?) but
the most commonly used name is the ENET, since the largest internal
use was within Engineering.
Now the whole company is being interconnected, and Engineering Network
is not really an appropriate name. But the E in ENET doesn't
necessarily have to stand for Engineering.
We think it can stand for Everthing, Employee, Everywhere, or whatever
anyone wants it to stand for.
The lack of any serious central control (other than a nodename
registry) makes things like this not really matter.
------------------------------
Date: 6 Jan 84 11:08:50 EST (Fri)
From: Chris Torek <chris%umcp-cs@CSNet-Relay>
Subject: Re: CSNet
CSNet ARPA relays are currently Rand-Relay and CSNet-Relay (not
UDel-Relay).
Chris
------------------------------
Date: 6 Jan 84 17:53:05 PST (Friday)
From: Jef Poskanzer <Poskanzer.PA@PARC-MAXC.ARPA>
Subject: Re: The Plethora of Networks
Here's a network you left out: the XEROX Internet. Most outsiders
tend to overlook the XEROX Internet, for various reasons:
o only a small proportion of the traffic is gatewayed to or from
other networks;
o what little gatewaying there is gets done almost invisibly;
o the name difficulty. (I'm told that XEROX used "Internet" first,
but that doesn't matter much now.)
The XEROX Internet only has about 2000 users, but it is widely
distributed, with users in Europe and Japan.
The mail transport mechanism within the XEROX Internet is called
Grapevine. Grapevine addresses look like "<user>.<registry>". If the
registry you're sending to is the one you are in, you can leave it
off, and the address becomes merely "<user>". Registries are
geographic - the two largest are "PA" (Palo Alto), for Northern
California, and "ES" (El Segundo), for Southern California.
To send mail in from the ARPAnet, the address looks like
"<user>.<registry>@PARC-MAXC". If the registry is PA, you can leave
it off, giving "<user>@PARC-MAXC". This is what I mean by invisible
gatewaying - to outsiders, it looks like all 2000 of us Xeroids
receive our mail on poor little PARC-MAXC. Not so - it's just a
gateway. I think the source of the confusion is that people are used
to explicitly specifying a host for the mail to be delivered to, as
well as a user on that host. Grapevine's mail servers are politely
invisible.
Sending mail out to the ARPAnet is as easy as pi. "ARPA" is just
another registry, so I just say "<user>@<host>.ARPA". Or if I'm
really lazy, I can just say "<user>@<host>", since anything with at
atsign automatically goes to the ARPAnet.
---
Jef
------------------------------
Date: Thursday, 5 Jan 1984 19:32:56-PST
From: Jim Burrows <decwrl!rhea!kryptn!ts1!burrows@Shasta>
Subject: Re: Dvorak keybords again (and again)
The following is a re-working of a message sent a few months ago to a
DEC in-house mailing list about the same topic. I have seen a couple
of other messages that get resubmitted again and again as the DVORAK
lengend is recounted.
--------------------------------
It's amazing how legends grow. There are two bits of misinformation in
the discussion of Dvorak keyboards, and one bit of information left
out.
First, the QWERTY (Sholes) keyboard was not the product of reverse
human engineering. Speed of typing was not the problem they were
trying to solve. Rather, they were trying to get keys that were typed
successively and fast to be as far away from each other as possible.
Therefore the layout was designed to seperate common letter pairs so
that the two letters were typed with different hands. Beyond that they
put the letters of the name of the device in the top row. There is no
evidence either that slowing down the rate of typing was the intent of
the design or that it was achieved.
Second, it was not mentioned that in the 50 years since the Dvorak
keyboard was created, no definitive experimental evidence of its
superiority has been established. There are about an equal number of
studies showing small improvements, and showing small degredations of
performance. Most studies have showed no signifigant difference.
However the keyboard layout has become a religious issue, and both
sides cast aspersions on the accuracy of the others tests. An
explanation of why the Dvorak layout doesn't work as well as you would
expect is the designed in pattern of alternating hands. The one thing
that has been demonstrated is that both alphabetically organized and
randomly organized keyboards are much worse than either QWERTY or
Dvorak.
Finally, the assertion that Touchtone (TM) keypads were intentionally
badly laid out is absolutely false. Bell undertook a massive effort to
layout the numeric pad, and produced one that is clearly and
demonstrably better than the adding machine layout used on many
numeric keypads, and to every other numeric layout tested. It is in
fact a classic study, and one of the examples of a company taking
great care to look before they lept.
I doubt I'll ever see the death of these bits of misinformation, but
as an active member of a human engineering research group I feel its
worth a try.
--------------------------------
I find it particularly interesting that the note about the telephone
keypad is appropriate to the Human-nets discussion as well as to the
original.
/s/ Jim Burrows
------------------------------
From: onyx!bob%amd70@BRL-BMD.ARPA
Date: 5 Jan 84 22:56:10 PST (Thu)
Subject: Re: HUMAN-NETS Digest V7 #1
The reason that the Touch Tone phone keyboard is the opposite
from calculators is that on The Touch Tone the keys are 1 - 9
followed by "O" for operator from left to right, top to bottom
the way ordinary citizens read. This was the reasoning of AT&T at
the time. They did consider making the Touch Tone keyboard the
same as business machines such as calculators.
Bob Toxen {ucbvax,decwrl,ihnp4,dual,fortune}!amd70!onyx!bob
Onyx Systems, San Jose CA
------------------------------
Date: 6 Jan 1984 0821-PST
From: SOROKA@USC-ECLC
Subject: telephone vs calculator keyboards
As I recall ...
When push-button telephones first appeared in the US, accountants
asked why they the keyboard differed from that of existing adding
machines. Calculators, which came later, simply took their keyboard
arrangement from that of the adding machine.
------------------------------
End of HUMAN-NETS Digest
************************