[ih] How long did it take?

[Jack Haverty]

Two additional historicak tidbits, from sometime in the 1980s:

1) Someone (can't remember who) told me that SS7 (Signaling System 7) 
used an embedded packet network to act as a control channel for doing 
its work.  That made sense as a way for AT&T/BellLabs to try out the "it 
can't work" packet technology as an internal mechanism to learn more 
about its reliability and how such newfangled technology behaved.

SS7 involved "out-of-band" control, where earlier mechanisms in the 
telephone network used "in-band" mechanisms.  The earlier in-band 
mechanisms were favorites of "hackers", since they involved simply 
sending the appropriate tone over the circuit to trigger some action (if 
curious, search for "Captain Crunch Whistle").  Ben Barker was the 
hardware engineer who set up the earliest IMPs and he told me that he 
had learned how to put a circuit into loopback by simply whistling into 
the modem handset.  Much easier than calling the circuit's tech support.

When TCP 2 was being evolved into TCP/IP4, there was a discussion about 
whether to use in-band or out-of-band for functions such as ICMP.  There 
are pros and cons of each but eventually the decision was made to use 
in-band signalling, which seemed easier than trying to get two disjoint 
networks to work at the same time.

2) An incident in the ARPANET indicated that other "packetization" 
schemes were likely in use internally within the "circuit switched" 
infrastructure:

At the time, circuits interconnecting IMPs generally came from ATT Long 
Lines.  Each IMP timing parameters had to be configured to work 
appropriately with either terrestrial or satellite-based circuits. The 
IMP software had been evolved so that IMPs could actually measure 
circuit behavior and configure itself.   The incident I remember was 
somewhere in the US involving a circuit between two IMPs.  The IMPs 
reported a problem with the circuit (somewhere in Georgia?).   The 
measurements didn't make sense to the IMP algorithm. IIRC, the measured 
delay was something like 90 msec.  That was far too low to be a 
satellite circuit, but also far too high for a traditional terrestrial 
circuit.

Someone at BBN investigated with the circuit provider (ATT IIRC). 
Apparently there was some kind of internal mechanism involving a 
microwave tower on top of a mountain, with a "beam" that would rotate 
around to send data to several other towers.    The data for "our" 
circuit was queued somewhere until the beam got to the appropriate 
position.  That explained the longer than usual delays for the 
traditional terrestrial circuit.  I don't know exactly how that scheme 
worked but it seemed that it had to involved splitting data streams up 
into short segments and queuing them until the right time to send.  
  Sounds like packet switching to me.

Sorry I can't nail down the timing any better than 1980s.   Perhaps 
AndyM remembers more..?

/Jack Haverty



On 9/17/26 17:20, the keyboard of geoff goodfellow via Internet-history 
wrote:
> On Wed, Sep 16, 2026 at 5:10 PM John Day <jeanjour at comcast.net> wrote:
>
>> Yes of course we know that. But that wasn’t what I asked.
>>
> ok, so how about this Grok conversational exchange then:
>
> TO GROK:
> *it's said that in the 60s, Paul Baran backed out of an offer from AT&T to
> build a packet switched network because he sensed wisely they didn’t
> believe in it and would not put the effort into to make it real.  any ideas
> of when did AT&T start to take packet switching seriously?*
>
> GROK REPLY:
> "The story is essentially right in spirit, though the mechanics were a bit
> different from “AT&T made him an offer and he walked away.”
>
> In the mid-1960s the Air Force was interested in Baran’s distributed
> “message-block” (packet) design. AT&T was the obvious builder. AT&T
> engineers rejected it on technical and institutional grounds—“it can’t
> work, and if it did we don’t want a competitor to ourselves”—and the
> Defense Communications Agency that would have overseen any implementation
> was staffed largely by people with analog-circuit backgrounds. Baran
> concluded they would either botch the project or quietly kill it, so he
> recommended that the whole effort be shelved rather than handed to them.
> The project was quietly dropped around 1966. ARPA later picked up the ideas
> (plus Donald Davies’s independent work) and built ARPANET without AT&T.
> When the Pentagon asked AT&T in the early 1970s whether it wanted to take
> over ARPANET and turn it into a public service, AT&T declined again. [
> https://www.nytimes.com/2011/03/28/technology/28baran.html]
>
> AT&T’s own conversion happened in stages:
>
>     - Late 1970s–early 1980s: Bell Labs (Sandy Fraser) developed Datakit, a
>     proprietary virtual-circuit packet switch. Internal test switches appeared
>     around 1979; it became an AT&T product in 1983 and was widely used inside
>     the Bell System for operations-support systems. This was the first time the
>     company treated packet switching as something it would actually build and
>     operate at scale. [https://www.youtube.com/watch?v=ojRtJ1U6Qzw]
>     - 1984: AT&T launched Accunet Packet Service, a public X.25 network
>     (tariff effective August 1984). That was its first commercial packet
>     offering to customers. [
>     https://bitsavers.trailing-edge.com/pdf/dataDecisions/1985_Communications_Systems/Vol2_810.pdf
>     ]
>     - Early 1990s: Frame Relay and ATM services followed.
>     - Late 1990s: AT&T installed its last new 4ESS circuit switch (1999) and
>     publicly described the future of the network as packet-based. Data traffic
>     overtook voice on the AT&T network around 2000 [
>     https://web.archive.org/web/20070107114157/http:/www.corp.att.com/history/nethistory/milestones.html
>     ]
>
>
> So the company began taking packet switching seriously for its own internal
> use in the late 1970s and for paying customers in 1984—roughly 15–20 years
> after Baran’s original proposal and a decade after Telenet and Tymnet had
> already commercialized it."
>
> ➔➔https://x.com/i/grok?conversation=2100738437875892246
>
> On Sep 16, 2026, at 20:05, the keyboard of geoff goodfellow <
>> geoff at iconia.com> wrote:
>>
>> Grok sez:
>>
>> AT&T operated X.25 packet-switched networks in the 1980s.
>>
>> The main one was ACCUNET Packet Service, a public data network that AT&T
>> launched as a regulated basic service. The FCC tariff took effect 18 August
>> 1984. By 1986 it had 99 service points in the U.S. Subscriber interfaces
>> followed CCITT X.25 (1980 recommendation, with a later move toward the 1984
>> version). It offered switched virtual calls and permanent virtual circuits
>> over access lines at 4.8, 9.6, and 56 kbps (via Dataphone Digital Service
>> or analog private lines). Nodes were interconnected with 56 kbps DDS
>> facilities.
>>
>> AT&T also ran a separate value-added X.25 service called Net 1000 (AIS*Net
>> 1000) through its unregulated subsidiary (American Bell / AT&T Information
>> Systems). That offering added protocol conversion, processing, and storage
>> on top of an underlying X.25 packet network supplied by the regulated side
>> of AT&T. It was withdrawn in 1986 after large losses.
>>
>> After the 1984 breakup, the long-distance AT&T entity continued ACCUNET
>> Packet Service while the Regional Bell Operating Companies offered their
>> own intra-LATA X.25 packet services (some of those RBOCs later became part
>> of today’s AT&T). Bell Labs also did packet-switching research and built
>> systems such as LADT that used X.25 internally.
>>
>> AT&T was late to commercial packet switching compared with Telenet and
>> Tymnet, but it did own and operate X.25 networks once it entered the market.
>>
>> ➔➔https://x.com/i/grok?conversation=2100374688069489036
>>
>>
>> On Wed, Sep 16, 2026 at 4:07 PM John Day via Internet-history <
>> internet-history at elists.isoc.org> wrote:
>>
>>> In the 60s, Baran backed out of an offer from AT&T to build a packet
>>> switched network because he sensed wisely they didn’t believe in it and
>>> would not put the effort into to make it real.
>>>
>>> We also all know the story from the ARPANET demo at ICCC72 when Bob
>>> Metcalfe was demoing the TIP to several AT&T attendees. The TIP crashed and
>>> re-booted from its neighbors. Bob thought this was a great aspect of the
>>> system. The AT&T guys thought it showed the ARPANET was a joke and not
>>> serious.
>>>
>>> So when did AT&T start to take packet switching seriously?
>>>
>>> Take care,
>>> John
>>> --
>>>
>>>

-------------- next part --------------
A non-text attachment was scrubbed...
Name: OpenPGP_signature.asc
Type: application/pgp-signature
Size: 665 bytes
Desc: OpenPGP digital signature
URL: <http://elists.isoc.org/pipermail/internet-history/attachments/20260917/902b04ff/attachment.asc>