[mailhist-discuss] Segments of email history

Dave Crocker dcrocker at gmail.com
Tue Apr 3 12:45:01 PDT 2012

On 3/31/2012 9:38 AM, Craig Partridge wrote:
> Seamless email relaying

This struck me as an interesting and unexpected item.

Plus I like having the lead being the summary of the innovation and I recommend 
it for all our discussions.  That is, for any candidate, we should lead with its 
value proposition.  What did it do that changed things?

This particular one hadn't occurred to me and yet it was of massive importance 
for infrastructure operation. It really came in two stages that probably should 
be marked separately:

      1. Seamless addressing, with the DNS.  The domain name system provided a 
naming overlay that became universal and supplanted local-part source-routing 
hacks of %, !, etc., for relaying across heterogeneous administrations and 
technologies.  (I could argue that the %-hack wasn't strictly source-routing, 
but it would be a very weak argument and it doesn't matter for this item...)

      2. Email routing, with the MX record.  Allowed remote sites to be as easy 
to contact as ones "directly" connected to the network  core.  As routing 
protocols go, this one is remarkably simplistic and yet it has proved 
sufficient.  And it took some iterations to  get right.

> Mark Horton's creation of uuencode/uudecode.

I'd label this "First de facto 'attachment' convention".  Was it in fact the 
first that gained widespread adoption?

On 3/31/2012 10:54 AM, Thomas Haigh wrote:
 > ·Pre-email electronic message transmission, including the suggestion that SAGE
 > operators could hack personal messages into data streams, etc.

We are going to need a solid definition of email, so that it marks a clear and 
reasonable boundary and makes it easy to exclude what came before.

 > ·Commercial offering of email by timesharing service

I've generally thought that onTyme (Tymenet's email service) was the first 
commercial email offering.  I think it was operational in the mid-70s! )

 > ·Integrated address book

Good point.  I don't remember whether Hermes had something.  I know that Eudora 
did, a bit later.

 > ·Confirmed delivery over network

I think this showed up in the LAN-based (single-server) systems before 
network-based system, but there was a network-based commercial enterprise system 
whose name I'm forgetting.

 > ·Email systems offered to smaller businesses by office automationvendors (DEC,
 > Wang, IBM, etc -- not clear what they had when in late 1970s)

And cc:mail and some others.  I'm unclear about the exact timeline, but suspect 
the LAN systems weren't significant until the mid-80s, whereas the others were 
really more mainframe-oriented and earlier.

 > ·Email service targeted at consumers

"targeted at consumers" could be tricky.  What does that mean, exactly.  Did 
compuserve qualify?  (e.g., it wasn't the best UI in the world...)

 > ·Rich text

oh boy.  this one is interesting but could be challenging.

 > ·Integrated encryption

This presumes we have it now...

 > It's possible that there are some areas in which Internet email does not claim
 > the "first."

Certain.  Not just possible.

The LAN systems were where most of the user-level innovation was taking place in 
the 80s, I believe.

On 4/2/2012 7:53 AM, John Vittal wrote:
 >> > * Header fields
 >> As best I can tell, this one was one of Ray Tomlinson's innovations.

It would be good to lock this down.  I hadn't realized that he created SNDMSG. 
And I hadn't realized that header fields came after its initial creation.

 >>> * Confirmed delivery over network
 >>> * Unread message recall capability
 > MSG had this probably as early as 1975. Hermes, also.

I don't recall (pun) MSG's having recall, as in deleting messages stored in a 
recipient's mailbox, nevermind on a remote machine.

I've always thought of this as a LAN-system feature that didn't generalize.  It 
mostly hinged on having a central, department-level database and a /highly/ 
integrated mail system for the department and/or enterprise.

 >> > * Client/server email, with messages downloaded for local processing,
 >>> filing, response on a different computer from the mailbox store.
 > Probably with the advent of POP in 1984?

Seems to be Laurel at Parc, before that?

No doubt there are some gradations of innovation here.

The UA/MTA model might fit in here, depending on the definition of 
client/server.  (That is, independent processes, versus independent machines.)

One can, I think, easily argue that /any/ user program that is distinct from a 
transfer program, qualifies under a basic definition.  In which case, 
sndmsg/cpynet probably qualifies.

To get to POP, we need a definition of intermediate delivery and then 
"offloading' to a personal or distinct computer, I think.

When I did the email architecture RFC, recently, the email community took the 
stance that POP and IMAP were equivalent, architecturally and that they came 
after delivery, per DSN behavior.  (I argued otherwise, but lost; I think POP is 
a delivery protocol.)

 >> > * GUI for client
 >> My bet is Novell or someone of that vintage.

Laurel predates that, doesn't it?

 >> I think Lotus Notes led here (though I remember that the Lotus guys
 >> struggled with email -- it didn't make sense to them).
 > Might one be able to argue that Vezza's MSGDMS or BBN's Hermes fit the bill here
 > before Lotus?

Almost certainly yes.


   Dave Crocker
   Brandenburg InternetWorking

More information about the mailhist-discuss mailing list