Ethernet names...

Tapley, Mark mtapley at swri.edu
Fri Oct 5 22:56:58 CDT 2018


> On Oct 4, 2018, at 11:00 PM, Eric Smith via cctalk <cctalk at classiccmp.org> wrote:
> 
> On Thu, Oct 4, 2018, 14:19 Eric Smith <spacewar at gmail.com> wrote:
> 
>> except that in the Pelkey account the Alto network wasn't designed and
>> built until June, _after_ the name change.
>> 
> 
> I should have written that it the design and construction _started_ in
> June. The initial Ethernet wasn't completed until late 1974.

Below, my interpretation of Dr. Metcalfe’s talk. Dates were generally backed up by pictures of documents, with dates visible on the covers, so I suspect the timelines given are reasonably accurate. I didn’t hear anything to contradict the timeline Grant was describing. The talk was fun to listen to, lots of dry humor and very good scene-setting and reminders of what the computational environment used to be like.

———

Metcalfe claimed he got his start at serial data transmission "Sending bits 1 at a time down an acoustic delay line memory”. The other parts of that project didn’t deliver, but he learned a lot from that.

He went on to connect a PDP-10 to the network at MIT, using similar serial technology.

He showed a document dated Oct. 1972, "Scenarios for using the Arpnet”

He got a Harvard PhD in 1972
He went to Xerox PARC in 1972

He showed a later photo of a Xerox Parc Alto personal computer, with Ethernet, and stated the intent there was to put that on the desk of all of the engineeris at PARC.

He started with 1/2" yellow coax - yellow because that was the color his group decided to order. He says that is the standard color, which standard is almost universally ignored.

He showed a diagram, including both the term "ether” - because the medimum could be coax cable - or telephone “ether", or radio “ether” - which presages the 802.11 standards. 

He showed a photo of an Aloha packet network radio system circa 1970. To the question, why did they not go not straight to wi-fi? He responded that RF technology would support no more than 4kbits at the time, while semiconductors developed in the 1990's allowed transition to radio at hundreds of kbits/s. They wanted at least hundreds of kbits/s, so they went with cable.

The first version worked at 2.94 Mbits/s, due to card space restrictions - a 170 ns clock present on the backplane of the Xerox Alto was the determinant, since there was no place to put a clock on the original card (which had to go into the Alto).

The bandwidth to his office went up by * 10,000 on installation of his card, vs. the old 300 bps modem it replaced.

He founded 3Com corporation in 1980, named for "Computer Communications Compatibility", intending to generate that compatibility via standards. He said he would not start a company name with a number if he had it to do over again.

Steve Jobs helped him start the company, invited him to the premier of Toy Story and sent a limo, and introduced him to Regis Mckenna (sic?) who was the marketing guy for 3Com for a few years. 

3Com’s first shipping hardware was a transceiver running at 10 Mbits/s, costing $750.

PC's at the time were not powerful enough to merit 10 Mbits/s, so 3Com pivoted to supply to Unix workstations, including Sun. By then they were moving to thin-net. 

He showed a draft Standard "blue book" Sept. 1980. Digital, Intel, Xerox were major participants.
 
Competition from IBM and GM (?) led to 3 competing standards: IEEE 802.3, 802.4, 802.5. 

He said 3Com shipped token ring devices before IBM did, but had "compatibility” issues in IBM environments due to IBM implementation. 

In Sept. 1982, 3Com shipped etherlink for IBM PC at $1000/card. 

He showed a graphic of a 3Com sales tool Circa 1982, showing a plot of the value of the network as being proportional to the number of “connections” which is the square of the number of nodes. This formula was described as "Metcalfe's law" in Forbes magazine in 1995. 

Part of 3Com’s success was its ability to demonstrate network effectiveness as it would be ~10 years in the future by looking at PARC, which his competitors could not do. He referred to this as a “time machine” in marketing research.

——————

	There was more to the talk, but that’s most of the historically relevant things he said.
	Hope this is helpful.
							- Mark



More information about the cctalk mailing list