Photos Spare Cycles MythBusters

Talk: Bob Metcalfe

Bob Metcalfe gave a short talk as part of the Ethernet celebration. It was a hurried, spirited talk off of notecards. My chaotic notes (along with some statements by Liddle) are below.


1982: there are people using ethernet i have never met
1986: there are people inventing ethernet i have never met

theme: conflict of ethernet

key ideas:
ether itself
business model

george santyana: those who don't recall the past are condemned to repeat it.
history is a pack of lies written by people who weren't there.

ethernet invented
tcp/ip invented in seminars by vince cerf at stanford

ethernet and tcp/ip evolved in lock step

bob metcalfe motto: "invention is flower, innovation is a weed"
- ethernet got annoying when it left parc

godzillas (too fast to transcribe them all)
- nova mca
- signet (simone infinitely glorious)
- quantum noise
- chuck thacker: xnet
- rs232c
- sneakernet
- alohanet
- hyperchannel
- 802.3
- gm token bus
- ibm token ring
- appletalk
- star lan
- cambridge ring
- atm
- and more

- sonet?

godzilla categories
- nova mca was first lan, could connect 15 mini-computers near each other
- signet, chuck thacker orignally working on it, turned over to metcalfe. charles went to bravo -> word
- chuck thacker started working on xnet. Metcalfe asked Bob Taylor to get Chuck to stop work on xnet
- quantum noise. ethernet was a bad idea b/c it was not "quantum noise limited"

outside of Xerox: attacked on the low-end
- arcnet. didn't submit as standard
- omninet: disk sharing lan. worked very poorly
- appletalk: worst of godzillas. 24kbs. Cheap and easy-to-use.

attack from within
- thought 10Mb was too expensive, wanted to stick with 2.94Mb ethernet
- Just after thick ethernet invented. Ron Crane: moved to thin ethernet. etherlink
- starlan: twisted pair. orignally 1Mb, went away when discovered 10Mb possible over twisted pair

0th idea: research is a good idea (cozy, rich, hot house)
1st idea: packets. took from alohanet.
2nd idea: protocol layering. allowed ethernet to be simple, unreliable.
3rd idea: distributed. put most on edges, less in middle. more scalable, reliable.
4th idea: ether itself. Originally Alto-Aloha Network. Key idea was that it would be medium-independent.
anticipated other media. planted the idea that ethernet would evolve.
5th idea: business model. Metcalfe claims this idea as the most robust.
- vertically integrated IBM biz model
- horizontal model: AOL, Cisco, Microsoft, Intel
- open source linux model: metcalfe disparages
- ethernet model: open standards, proprietary implementations. must be interoperable. high value on maintaining installed base.

Three new godzillas:
something comes along and beats ethernet, gets called ethernet, so ethernet can't lose

fibre channel: 10gig ethernet will kill it
sonet: T1 is doomed
cellular telephone: 802.11 will beat it

david liddle:

his last intellectual contribution to ethernet was pointing boggs at cable parts guide for vampire taps

villians of ethernet
- ibm wanted token ring b/c of 3270 (which has 32 bytes of memory and needed to be "fed" often)
- hp's hpiv had just been IEEE standardized. no real technical problem with ethernet.
- argued about "finite addressing".
- liddle: enough to address every kilogram of matter in the universe.
- hp executive: we'll soon be making computers less than 1kg.
- liddle: so are you going to use all the matter in the universe?

- the thing that pushed ethernet through IEEE was giving the spec to ECMA. IEEE argued about bit placement just to not agree (hp wanted their fingerprints on it). Once they saw that it could become an ECMA standard faster, it became an IEEE standard.

- ethernet licenses were licensed cheaply, but license required absolute compatibility


related entries.

what is this?

This page contains a single entry from kwc blog posted on May 22, 2003 12:38 AM.

The previous post was William Gibson on the progression of media.

The next post is Go Annika.

Current entries can be found on the main page.