[ih] The Atlantic on Email

[Dave Crocker]

On 3/14/2016 8:14 PM, John Klensin wrote:
> The article is fairly badly damaged in other ways.  For example, it
> suggests that people tried "!" and "%" instead of the "@", implying
> that both were used in Internet mail.  Of course, "!" was used in uucp
> routing "bang paths" (not on the Internet except as the
> pseudo-local-part associated with a gateway) and, at least AFAIK, "%"
> was used in local-part routing, always in conjunction with "@".

I reacted to that bit of the article, too, but decided that it suffered 
more from indulging in bit too much drama than in outright error.

Arguably, uucp source routing was initially an alternative to arpanet 
addressing.  Hence, ! was an 'alternative'.

The use of % you cite, as coupled with @, was indeed what I set up for 
CSNet and was always fully coupled with Arpanet/Internet mail.

However I seem to recall % also being used for source-routing on Bitnet, 
which was originally completely independent of uucp mail or arpanet mail.

Convergence on @ and domain names, with a common addressing scheme, was 
an explicit effort among the disparate email communities, in the 
mid-1980s.  I'm told this included heated debate...


> Once
> it took off, "@" was fairly universally used: the exceptions were
> Multics (and possibly a few other systems) where the symbol turned out
> to be used as a line-kill character and on some systems where "@" was
> hard or impossible to type (e.g., early version of ISO 646 defined

That's why RFC733 accepted " at " as an alternative to @.


> Beyond that, depending on how "real Internet" is defined, it was
> fairly well deployed before BITNET/NETNORTH/EARN really got going.

For a mass-market consideration of "Internet", perhaps.  For this 
discussion group, I'd say no.

Here, Arpanet was before them, but Internet was not yet well deployed 
before the others were well underway.  But perhaps only be one or a few 
years.


> The other strangeness of the discussion is that, if my memory is
> correct, inter-machine email predates ARPANET/Internet email.  The
> example I'm thinking of wasn't very fast because the transport
> mechanism involved mag tape over sneakernet, but it was inter-machine,
> demonstrating that one needs to be very careful about what is being
> claimed.  What I don't remember was whether there was a way to
> designate the other machine or whether that was just deduced from
> accounting information.  Tom Van Vleck would presumably remember if
> anyone thinks it is important.

I have never seen anyone previously claim that sneakernet-based file 
exchanges qualified as inter-machine email, nor that there was an 
addressing scheme to support referencing such an exchange.

These are the two innovations specifically credited to Tomlinson.  And 
always have been, that I have ever seen.

d/

-- 

   Dave Crocker
   Brandenburg InternetWorking
   bbiw.net