[ih] Email and address books

[Marty Lyons]

On May 23, 2012, at 10:03 PM, Alan Maitland wrote:
> 
> Not sure if that counts for this discussion and I'm guessing there was probably some equivalent in the DEC and IBM environments, though I don't absolutely know that to be true.  If my memory is correct, these things were around in the circa 1986 timeframe, so perhaps they predated Eudora.


Mail aliases existed through various hacks on Unix for a very long time; I remember using aliases in 1981.  I could probably dig up my old files if I looked hard enough.  There was a lot going on in the UUCP world around that time.  There was later a wide variety of user agents beyond just mail, mailx, and the emacs world; I think mh goes back maybe the earliest (1977) [1, p. 5, footnote 2], and support for aliases [1, p. 6].  In 1986 elm as an MUA on Unix also had good support for aliases. 

In those early days, actually getting mail to where it was intended to go in not on ARPAnet was a challenge; using a mail agent alias file was critical to keeping track of how to get a message to someone.

For sites connected by UUCP,  the UUCP Mapping project used site-prepared files which could be fed into pathalias to produce automated email path routing.  I maintained one of the early gateway maps showing how to transit from one net to another (e.g. from UUCP-land to BITNET, ARPA to IBM, CSNET to BITNET, and so on)... we had to keep track of all of this stuff manually, and the state of interconnections was changing all the time.   The big relay sites that were on multiple networks never "officially" gateway'ed mail, so we ended up with convoluted bang and % routing hack address, e.g.  hosta!hostb!user%decwrl.csnet at mit-multics.arpa.  Depending on whatever MTA picked up the file would parse the headers and hopefully do something intelligent for the next hop or final delivery.  RFC 976 (Mark Horton, Bell Labs, 1986) has some good notes on UUCP email header decoding for routing [2].  But once you figured out a way to reach someone, you made an alias, since most of the time it was trial and error and you didn't want to figure it all out again!

Unix sendmail configuration files around this time started to get even more bizarre than normal to handle the strange paths in mail headers.  There was eventually a way to deal with network gateways by using a separate file which was sent around UUCP... I can't recall what it was called or who maintained it.  Sometime around 1992, Zmailer emerged as a sendmail alternative to try to make email address routing easier at the transport level; recall even in the early '90s we were still in the post-NSFnet days and commercial networking was still in its infancy... some sites were still reachable only through UUCP using the % routing hack in addresses.
 
[I really should write all this up.  Looking back on what we did in the early days of interconnecting the disparate networks involved lots of hacks and heavy lifting all around.  For those of us not directly connected to ARPAnet, life was filled with work-arounds.]

In the BITNET world circa 1984, sites running IBM VM used the VM NOTE [3] command for mail, which stored aliases in a file called USERID NAMES.  The format for an entry was:

:nick.JOEUSER
               :userid.JOE
               :node.NODENAME.ARPA
               :name.Joe User
               :addr.Campus Drive;Big University;CA

"node" in the BITNET context had a particular meaning, but by inserting a properly configured address (with gateway), you had some hope of getting mail out of BITNET and on to another part of what was just starting to be called the "Internet" by routing mail out using "well-connected nodes".  In the BITNET universe that meant University of Wisconsin, Princeton, and the City University of New York.

Ultimately though on the BITNET side tools started being developed to conform to RFC822 -- Alan Crosswell (and others) at Columbia University wrote MAILER, which was a smart MTA for VM systems, allowing true routing over BITNET.  This was a *big deal* at the time, since nominally there was no way to send outside the BITNET domain.  MAILER was for BITNET nodes what sendmail was for Unix.  And on the user agent side, Richard Schafer at Rice University wrote the MAIL program, which did proper formatting, headers, and so on.  

IBM also sold a supported product called "PROFS" (Professional Office System) for VM which had it's own nickname file as well, that was edited from inside the application.  In the corporate world, PROFS was a big seller, but the university world on BITNET was running MAILER/MAIL.  I seem to recall coming up with some hack to actually gateway mail out of PROFS to the net, but I'm not sure I have those files any more. [Trivia: PROFS was originally written by ARCO in Dallas for their own in-house use.]

IBM systems finally became a first-class network citizen around 1988, when IBM started shipping the 8232 channel attached controller.  It was a $40,000 IBM microchannel PC with microcode to talk to a bus/tag interface on one side (that was, presumably, the $39,000 interface), and a Ethernet interface in another slot ($1).  The TCP/IP software product was IBM P/N 5798-FAL (I seem to remember the first generation had a different P/N, but everyone remembers the -FAL version).  With the introduction of the hardware and software, you could directly connect a VM host to an IP network directly. Almost immediately, Bus-Tech Inc., shipped a competitive device at much lower cost and higher performance, it used the IBM FAL code on the host.  Several years later, IBM added TCP/IP to the main lineup of communications controllers such as the 3745.

/Marty

P.S.  I just realized I'm retelling some old history, I wrote about this a few years ago:

http://mailman.postel.org/pipermail/internet-history/2010-January/001200.html

[1] http://www.rand.org/content/dam/rand/pubs/notes/2009/N3017.pdf
[2] http://tools.ietf.org/html/rfc976
[3] http://publib.boulder.ibm.com/infocenter/zvm/v6r1/index.jsp?topic=/com.ibm.zvm.v610.dmsb4/hcsd8c0175.htm