[TUHS] UUNET

[Clem Cole via TUHS]

Thanks, Rick - much of what I started to write, you have in Peter's
document.  But let me add a little more to it that he left out, before and
a bit after the UUNET experiment.


   - In the late 1979/early 1980 timeframe at the Winter Usenix in Colorado [we
   were in a rented movie theatre - that was featuring the new Sci-Fi
   Thriller: "The Black Hole"].   Steve delivered his first report on the
   UNC/Duke link.
   - One of the more amusing things that I remember about that was that we
   asked him how he was dialing.   Remember the funny Hayes "in band
   signaling" ( the later industry standard AT cruft) has not been invented.
    Also, remember that Unix is from TPC (The Telephone Company).  So, to
   have a computer source a phone call using the Bell System's "Direct
   Distance Dialing Network", you needed what was called an "Automatic Call
   Unit" (ACU), which the WE 801 was pretty much the TelCo standard, and DEC
   made the DN11, which provided the required RS-366 [
   https://en.wikipedia.org/wiki/RS-366 ] interface to the 801.
   - So UUCP, of course, assumes each PDP-11 (or later Vax) that wants to
   dial another system uses a DN11 and an 801 (and compatible WE212
   modems). Hopefully, this picture from the "DEC-11-HDNAA-B-D DN11 Automatic
   Calling Unit Interface Manual" - will pass through [if not, search for the
   PDF and its page 2-1]

[image: ACU-Picture.png]

   - WE212s and 801s were fairly expensive.  An many/most Universities
   tended to use Vadic VA3467 or VA3480 modems anyway.  But a few places did
   have Vadic's 801 equivalent, the VA811 installed in that modem rack to
   dial the phone lines for the modems.
   - Lacking an ACU, Steve and Tom, like all good phone system hackers of
   that time (*mei culpa*),  took a DR-11C [DEC's programmed I/O
16-bit parallel
   unibus interface] and connected a relay to it.  They then connected the
   "business side" of the relay across the TIP and RING wires of each phone
   line, so when the relay closed, it shorted TIP and RING together.   They
   knew that when you used a dial telephone, the pulses for each digit
   generated by the "finger wheel" on a traditional handset were a 40 ms
   "make" (close), followed by a 60 ms "Break" (open), with another
   "inter-digit pause" of 800 ms between numbers being dialed.
   - So they replaced the UUCP code to talk to DN11, to talk to their DR-11C
   dialer hack.
   - I know both decvax and teklabs 11/70 had Vadic gear with a DN11s and a
   proper ACU, but I'm not sure how many other sites use the Bellevin hack (it
   was cute, just lucky the local phone company never found out).



   - An important point to remember about this time is that those who had
   experienced ARPANET access, particularly people like myself who no longer
   had it, knew the power of a widespread mail system.
   - While Peter does a good job of explaining how *news took off*, the
   first 50-100 UUCP sites outside of the Bell System were *created just to
   move e-mail*.   Except for a very small number of places, connecting to
   the ARPANet was "unobtainium."   Even at UCB, ucbvax was not on the
   ARPANET; Ing70 was (their connection was a 960-cps serial link called
   Berknet). While decvax could connect via DECnet to DEC's PDP-20 (which was
   on the ARPAnet), mail could not flow via that link.
   - So after hearing Steve's talk, I came back to Oregon, and personally
   set up the teklabs to ucbvax link, since we had a number of students
   from the UCB CAD group (which I would later join) working in Teklabs.  I
   also set up teklabs to decvax, ihnp4, as well as what would become known
   as the "marx" machines that Brian Redman had in Whippany.  I was hardly
   unique.  I think many UNIX folks left that USENIX with the same idea.   I
   know Armando set up a number of uucp connections to different sites
   after that.
   - But that was the point; *we just wanted an email connection.*



   - Another point, if you were coming from the ARPANET world, you had
   likely experienced mailing lists like scify-lovers or risks-digest.  So
   when we heard about the Duke/UNC news system, a lot of us thought we could
   recreate that experience [little did we know what would happen].  So within
   a year or so after that USENIX talk, those same 50-100 sites started to
   exchange "news" also - it was only about 25 articles a day, but a few
   e-mail messages.   120 cps WE212-style modems were more than adequate.



   - For the first few years, the ARPAnet and USENET basically existed
   without each other.  But at some point, a couple of places, such as UC
   Berkeley, started to forward email between the two networks.
   - I do remember there was pushback from some of the ARPANET sites.  In
   fact, the 1981 creation of "CSNET" was part of that, and CSNET had its
   own UUCP-like scheme called PhoneNet.
   - By 1982, BBN ran a mail router, and CSNET at some sites (particularly
   those that were not ARPANET-connected) had both PhoneNET and UUCP
   connections.
   - Also, MaryAnn and Matt Glickman had released BNEWS, which had become
   what we all called "net news," and was now using way more bandwidth than
   E-MAIL did.
   - Another important thing happened in 1982.  Concord Data Systems and
   Racal-Vadic introduced the first modems to break the 1200-baud limit of
   POTS (their solution would become the V.22bis standard).  They did this,
   actually sending data at 600 baud, but since " baud " does not mean
   "bits," it means "information", they managed to send 4 bits in each baud
   and added error correction.   Plus, since these modems, unlike their
   earlier purely analog devices, had a microprocessor inside, they could do
   "smart" things.



   - As Peter tells you, by 1984, the amount of news traffic had started to get
   bad.  Even with the new Vadic modems that are twice as fast (240 cps),
   the transmission time was still way too long (I'll come back to modems
   in a bit).
   - Peter mentions, but really does not explain, the big ugly truth of the
   time.  A couple of very large sites (decvax, ihnp4, and seismo)  were
   handling the long-distance load.
   - When Lauren Weinstien ran his experiment, Armando had to report to a
   few of us what was happening at decvax.  It turns out the way DEC
   handled telephones was like heat, a utility for each site.  So the
   actual phone bills for decvax, which was the central server for the DEC
   Telephone Industry Group (which had been created years earlier to serve
   DEC's #1 customer, the Bell System), were never directly examined.
   Armando's boss, Bill Munson, told us that he had looked into it at one
   point and estimated that decvax was running about $500K/yr in phone charges,
   and at that point, when it was not being directly noticed, he stopped
   asking, not wanting to draw more attention to it.
   - I don't remember who it was, it may have been Mary Ann, but it turns
   out somebody at AT&T did look into the ihnp4 charges, but then
   calculated that for every phone call it made, they generated between
   10-20 downstream, so it was actually a good thing for AT&T.
   - I was running teklabs, and while we had some long-distance UUCP
   connections to places like ucbvax and decvax, we did not exchange news
   directly with either, only email.  I don't remember now who we got our
   primary feed from, but ISTR we sent news downstream only to reasonably
   local sites, such as Reed College and Portland State.
   - BTW: By now, UC Berkeley had become an informal gateway between the
   two networks.  IIRC, they had very few outbound UUCP connections.  But
   sites like decvax and teklabs called them.
   - Rick Adams was working within the "Center for Seismic Studies," which
   was part of a DoD contractor [I believe it was Science Applications
   International Corporation (SAIC), but I do not know if I have that right].
   - Their work was global, so his team had already incurred heavy
   telecommunications charges. But I have no idea how much they were
   spending.  And it was also through this work that he became familiar
   with Tymnet and CompuServe
   - Tymnet was a private, commercial, packet-switched data network that
   operated as a public utility, connecting remote computer terminals to
   central mainframe systems [I'm going to ignore a ton of technical details
   here that aren't pertinent to USENET history].  They set up what they
   called Tymsats, which were terminal interface processors like the
   ARPANET TACs around the world.     They supplied what they called Tycoms or
   host interface processors (think the ARPANET IMP) that interfaced to the
   major players' equipment (DEC, IBM, Burroughs, and the like).   Someone
   like SAIC obtained a Tycom and could attach their computers (primarily
   mainframes).  SAIC's customers, like the ones Rick was working with a
   seismo, could dial into local numbers and access the remote (SAIC)
   mainframe without incurring long-distance charges.  I don't know
the fee structure,
   as I was never involved with that side of it, but it was significantly
   cheaper (more in a minute).
   - Similarly, CompuServer worked like Tymnet; the difference was that,
   instead of a Tycom, they used X.25 PAD, and instead of Tymsats, they had a
   custom PDP-11-based system that did the same thing but ran their SW,
   which they called Nodes and T-Nodes.
   - UUNET would eventually use both services.



   - So the USENIX membership is now a little concerned.  We knew that at some
   point, the gravy train of decvax, ihnp4, *etc.* could not last —
   particularly at the rate that the amount of data news was creating daily
   was getting extremely large (megabytes).  At 120 cps, that was almost 2
   hrs and 30 minutes per megabyte, and that does not include any overhead,
   so it ended up being closer to 3 hrs.  Even at 240 cps, you were looking
   at about 90 minutes.



   - Coming back to USENET itself, by the summer of 1984, news delivery had
   become a huge issue.  Peter described Lauren's solution.  He came to the
   USENIX BOD with a proposal, which the BOD eventually backed.   I don't
   remember why it never really caught on, other than that decoding the
   transmission required some specialized hardware, which a few people were
   comfortable with.  Lauren, who worked in LA with broadcasters and other
   content creators, and a few others were, but I was at Masscomp by then, and
   when I looked into it, it was not going to be easy.   And the other issue
   was that Lauren solved the broadcast issue, but we still did not have a way
   to get replies sent back up the chain.
   - So the membership is a bit of a buzz, and the board was aware that we
   were all worried.   The good news for all of us at the time was that
   because of the "UNIX Wars," USENIX was fairly flush with cash from its
   conferences.
   - I don't know if Rick had siesmo on either Tymnet or CompuServe yet,
   but he did understand the cost structure and had a good idea of how it
   might work when he mentioned his idea to a few people in the summer of 1985.
   - It's interesting that the Telebit Trailblazer also came out in 1985.
   They quickly became the UUCP standard, because they were able to send 1800
   cps using a radical, proprietary modulation scheme called Packetized
   Ensemble Protocol (PEP). Rather than relying on a single carrier frequency,
   as standard modems do, PEP dynamically splits the phone line into up to 512
   separate micro-channels. The modem constantly analyzed the telephone line,
   automatically disabling individual micro-channels affected by static or
   line noise while keeping the clear channels wide open.  But its designers
   were really smart and embedded Greg Chesson's "g" protocol (the UUCP
   standard) into the firmware.   Each side of the UUCP link thought it was
   running the transfer protocol (without any modifications), and then the
   Trailblazer created an error-free connection over a standard POTS line.
   - So starting in 1985, if you administered a site serving the USENET via
   UUCP, you tended to run at least one Trailblazer.



   - As Peter says, in late May/June "*;login*" the BOD started asking
   people for proposals, which Rick replied to first in the fall, and by the
   Spring had put together a reason business plan.  But I should note that
   between his original October 1986 BOD presentation and the BOD's
   insistence, one of the things the BOD insisted on was that there had to
   be a reasonable plan for self-sufficiency.   The Association wanted to
   find a real solution and was willing to use its resources to help to
   seed a solution to the UUCP/USENET problem, but it could not be responsible
   for the solution itself.  The BOD didn't want the Association to be seen as
   providing any direct service that took away from its mission of creating
   a place for the community to come together.
   - I probably should mention that at that time. There was consideration
   of how it was different by providing a place to come together
   electronically, like news, different from putting on conferences and
   publishing proceedings and papers?  Again, this is before things like
   what we know today as social media. I think the board really felt that the
   optics of being a place that could support someone/head start things (like
   the mapping project, helping to fund "FOSS" versions of both pax and
   terminfo, etc.) were in their charter. But how far out of that
   could/should it reasonably go?
   - The idea of backing Rick and cosigning/being a financial guarantor was
   a bold step.   Remember that Rick needed to negotiate a data transit deal
   with someone (CompuServe was first), and because the monthly data traffic
   costs were projected to be incredibly high, CompuServe required a financial
   safety net.  The Association, which had a few million in assets by then,
   was that guarantor.   Frankly, that action did worry some folks on the BOD.



   - What started all this was a question about it being for-profit.  To
   the Association, it was an experiment to see if this would pan out.   While
   they were the guarantor, they wanted a legal separation, so UUNET also set
   up a non-profit as a separate entity.   The hope was that it would pay for
   itself, but some feared that it might not.     So, in two years, when
   UUNET was more than self-sufficient, it was time for the Association and
   UUNET to cut ties.    I think most of us were thrilled that Rick
had succeeded.
   The community had a real solution that seemed to have worked.



   - One thing I slightly disagree with in Peter's piece is his claim that
   Rick "created a service to fill the need" for anyone to access USENET.  The
   real driver, certainly as far as the Association was concerned, was how
   USENET would survive if the few large sites stopped passing data for free.


   - Rick is right that by the mid-1980s, finding a cooperative site to give
   you a news feed started to get harder.
   - From that standpoint, I had become an independent consultant around
   that time. I actually had a local news feed, but I still wanted to call
   into the local Tymnet Tysat with an unlimited access POTS line and my own
   Telebit to connect to UUNET, so email to me was never more than one hop.
   My business cards said:  ...uunet!ccc!clemc [which morphed into
   clemc at ccc.com when we finally combined the two separate namespaces].  My
   fees to Tymnet and to Rick were small enough that I could handle them —
   this was exactly the kind of outcome the Association had hoped to
   obtain.  Any site could be just one away from a major hub; the cost to do
   so was reasonable, and it was working.



   - To Rick's credit, he did not stop with UUCP.   He created AlterNet as
   the first ISP was coming to bear.   The service he offered was the same
   core idea, just a new protocol.  He turned a number of his original UUCP
   connections to be AlterNet soon thereafter.




On Thu, May 28, 2026 at 2:00 PM Rik Farrow via TUHS <tuhs at tuhs.org> wrote:

> I wondered about this, and based on communication with Rick Adams, Peter
> Salus has the best description of the evolution of UUnet:
>
>
> https://www.usenix.org/system/files/login/articles/login_aug15_09_salus.pdf#:~:text=First%20designed%20to%20operate%20over,to%20a%20need%20for%20improvements
> .
>
> Adams asked for seed money, used to set up a gateway to the ARPAnet so that
> people could send email once they moved on from universities that had
> access. The UUnet portion of this article begins on page 41, column 2.
>
> Rik
>
>
> On Wed, May 27, 2026 at 4:03 PM Kevin Bowling via TUHS <tuhs at tuhs.org>
> wrote:
>
> > https://en.wikipedia.org/wiki/UUNET
> >
> > Does someone (maybe Rich Salz?) know how it transformed from a USENIX
> > infused non-profit to a for-profit?  Controversial at the time?
> >
>