Interesting People mailing list archives
Internet still reshaping history
From: David Farber <dave () farber net>
Date: Wed, 9 Sep 2009 08:06:14 -0400
Begin forwarded message: From: Richard Bennett <richard () bennett com> Date: September 8, 2009 9:57:08 PM EDT To: dave () farber net Cc: ip <ip () v2 listbox com> Subject: Re: [IP] Internet still reshaping historyI have to take issue with some of Vint's recollections, based on documented historical accounts and direct contact with other parties involved. First one correction to my previous comment: the IMP never was an IP router, as I erroneously claimed; it was used, at some point, as an "IP relay" that carried TCP/IP packets over the ARPA subnet. The IMP code wasn't modified for this application, it simply treated IP as payload, much as Ethernet and Sonet do today.
I didn't say that Vint copied CYCLADES, I said the *architecture* that Pouzin et. al. developed for CYCLADES was used by the datagram networks that followed him, including TCP/IP and DECnet; not just sliding windows, but the notion of a sliding window, serializing, virtual circuit transport protocol running over a datagram protocol at the network layer. Arguably, it actually might have been better for the Internet if they had simply used TS/Cigale verbatim instead of going with something of their own invention: We would have had a world- wide network with interoperable protocols much sooner, and we wouldn't have all these addressing-based anomalies on the Internet. CYCLADES, DECnet, XNS, and ISO all gave the host an address apart from the addresses of its subnetwork points of attachment; TCP/IP doesn't, and that causes a number of problems with multi-homing and mobility that are causing BGP table size to explode. CYCLADES, DECnet, and the others were also capable of handling congestion without going into meltdown. TCP/IP's unique solutions to the addressing and congestion problems are proof that it wasn't a copy of TS/Cigale.
The cooperation of BBN and IRIA took place pursuant to an actual contract. If Vint doesn't remember it, that's of no particular significance since he wasn't a party to it. People from both the IRIA and BBN teams have confirmed that they worked together on the CYCLADES architecture.
Regarding 1983, my point is simply that many sites were already running TCP long before NCP was finally shut down. And yes, implementing TCP and IP in a number of host operating systems was hard work; there was a panel discussion on it at Sigcomm in about 82 or so that I attended, with Bill Joy and some others talking about how hard it was to implement some of the features TCP and IP demanded, such as checksums and byte streams and MBUFs. My only issue is that by the time NCP was officially dead, it had already been dwindling for quite some time.
I notice that the "official" history credits Kleinrock with inventing packet-switching; I won't comment further on this claim except to say that people who care should read Davies' paper "An Historical Study of the Beginnings of Packet Switching" (The Computer Journal, Vol. 44, No. 3, British Computer Society, 2001). It's all there.
RB Dave Farber wrote:
Begin forwarded message:From: Lauren Weinstein <lauren () vortex com> Date: September 8, 2009 18:26:51 EDT To: dave () farber net Subject: [IP] Re: Internet still reshaping history Dave,I passed along some of the ongoing Internet History discussion to Vint,who sent me back some notes for forwarding back to you and IP. First, he noted that:The MOST ATM system did trademark the term "Internet" and it took about10 years of negotiation (by CNRI - Bob Kahn's company) to undo thisaction. The trademark filing took place well after the Internet was inoperation. vint And: ----- Forwarded message from Vint Cerf <vint () google com> ----- Date: Tue, 8 Sep 2009 17:45:03 -0400 From: Vint Cerf <vint () google com>Subject: Re: [dave () farber net: [IP] Re: Internet still reshaping history]To: Lauren Weinstein <lauren () vortex com>Richard Bennett makes several inferences that I consider to be incorrect.Kindly see http://www.isoc.org/internet/history/brief.shtmlfor as concise a history as any I have seen by the people who actuallydeveloped both the ARPANET and the Internet. I used some of Louis Pouzin's ideas (especially the sliding window) for TCP flow control. One of Louis' colleagues, Gerard LeLann spent a year at my Stanford Lab in 1974 while the TCP was being developed. But I can categorically say that TCP was not TS and that we adopted different solutions to some problems while adopting similar ones such as a datagram substrate. I don't think it accurate to suggest that we simply copied either Cigale, Cyclades or TS. We certainly wanted to acknowledge Louis' influence on our thinking and we have regularly cited his, Gerard LeLann's and Hubert Zimmermann's contributions. We separated TCP and IP in 1977 to support real-time applications. I do not recall a close interaction between BBN and IRIA. The ARPANET/ALOHA net link was terminal level: ALOHA net connected terminals to the Univ Hawaii host via the MENEHUNE (essentially a terminalcontroller). Terminals then went, via the University of Hawaii host intothe Internet. The "terminal to host" ALOHANET was not carrying IP packets at all. I would not characterize this as a gateway of the IP form.Craig Partridge has already pointed out that TCP/IP was never in the IMPS,only in hosts and gateways. The ARPANET was untouched as it became a part of the Internet.Kleinrock's work established the mathematical basis for analyzing packetand message switched networks. Paul Baran's work was speculative buthe never got to implement it. Donald Davies coined the term "packet" andimplemented a one-node system at the National Physical Laboratory in Teddington, UK. The detailed IMP function was designed by BBN, notably includingBob Kahn, Dave Walden and others. Larry Roberts established the overallobjectives and much of the conceptual basis for the system. I strongly disagree with Bennett's characterization of Jan 1, 1983.A huge effort was put into developing TCP/IP for many operating systemsand for implementing and testing these in the live ARPANET leading upto shutting down NCP. That date also coincided with the splitting of theARPANET into ARPANET (research) and MILNET as well as formally including the mobile packer radio net in the SF Bay are, the Atlantic PacketSatellite network, networks in Europe, and ethernets at PARC and elsewhereinto the Internet environment. Until January 1, 1983, NCP and the applicationsrunning over it were always a backup but TCP/IP become the sole mode ofoperation after that date. Vint Cerf On Sep 8, 2009, at 11:44 AM, Lauren Weinstein wrote:----- Forwarded message from David Farber <dave () farber net> ----- Date: Tue, 8 Sep 2009 07:40:00 -0400 From: David Farber <dave () farber net> Subject: [IP] Re: Internet still reshaping history Reply-To: dave () farber net To: ip <ip () v2 listbox com> Begin forwarded message: From: Richard Bennett <richard () bennett com> Date: September 7, 2009 8:05:25 PM EDT To: Dave Farber <dave () farber net> Subject: Re: [IP] Internet still reshaping history Dave - for IP if you wish.The establishment view on the origin of the Internet is not correct. Ifweunderstand the Internet as a network with a virtual circuit transport protocol layered on top of a datagram protocol at the network layer, itwasreduced to practice in France in the CYCLADES network in 1973 (see: The CYCLADES Computer Network, L. Pouzin co-author and ed., North Holland1982, ISBN: 0 444 86482 2.) The transport protocol was called TS and the datagramprotocol was called CIGALE. CYCLADES covered all the research centers inFrance and was ultimately connected to the UK and other points in Europe. Wisely, Cerf appropriated the architecture of CYCLADES, as did thedesigners of DECnet. And Cerf makes no bones about it, as you'll see inthereferences to Louis Pouzin, the inventor of CYCLADES, other members ofhisteam, and common advisers at BBN in Cerf and Kahn's first paper on TCPin1974 and the first TCP RFC, co-authored by Cerf, Sunshine, and Dalal,also in 1974. By the 1980s, Internet people mysteriously stopped giving credit to CYCLADES. For the timeline, TCP and IP weren't even separate entities on paper until what, 1975-6? And in code somewhat later...So it was actually the team at IRIA in France, headed by Louis Pouzinand working closely with BBN, who first reduced internetworking to practice.That is, unless you don't count the gateway between ARPANET and ALOHA asanexample of internetworking. Xerox also did some interesting things with datagram protocols in this era, PUP, but their lawyers didn't want toomany people knowing what they were doing. I think the idea that the ARPANET was an early-stage Internet comes about because the hardware and comms infrastructure that made ARPANET work waslater appropriated by the TCP/IP Internet. In some places, it was just asoftware upgrade that turned an ARPANET IMP into an IP router, and people continued to use the name "ARPANET" for many years after TCP came along.These anniversary stories (several are circulating now, based on an AP story) are pretty aggressive in promoting the claim that Len Kleinrockinvented packet-switching, the ARPANET, and the Internet. Old-timers areaware that there's a lot of controversy around these claims, some of itcaptured in the Donald Davies paper He's pretty persuasive that the honors forpacket-switching belong to Paul Baran and himself. Kleinrock did veryinteresting work on queuing theory, but that's not packet-switching. There's also a suggestion that the folks at UCLA (or the NWG in general)built the ARPANET, but we know that the code in the IMP was all writtenbythe BBN folks, Dave Walden and colleagues, and the NWG contribution was limited to the Host-Host protocol. The plan for the ARPANET, functionsof the IMP, deployment, pricing, etc, was done by Larry Roberts, who presented it at a very interesting ACM conference in Gatlinburg, TN, in 1967; See: Lawrence G. Roberts, “Multiple Computer Networks and Intercomputer Communication,” ACM Symposium on Operating Systems Principles, Proceedings of the first ACM symposium on Operating System Principles, New York: ACM, 1967. Pg 3.1-3.6.The thing that showed up in the UCLA lab in 1969 is what Roberts calls "ARPA II;" ARPA I was a packet-based timesharing system construct - a paper network - devised by J. C. R. Licklider in 1962-64. See: LawrenceG.Roberts, "The evolution of packet switching," Proceedings of the IEEE , vol.66, no.11, pp. 1307-1313, Nov. 1978. Licklider's theoretical workinspired Roberts to build ARPA II and Davies to build the packet network at NPL in the UK. The Jan 1, 1983 event wasn't all that significant since much of theARPANET had already converted to TCP/IP; this was simply the date whenNCP was turned off (although it wasn't; like all schedules, it slipped.) Just a few random thoughts I had handy on this fine holiday. RB David Farber wrote:Begin forwarded message: From: "David P. Reed" <dpreed () reed com> Date: September 6, 2009 9:18:35 PM EDT To: dave () farber net Cc: ip <ip () v2 listbox com> Subject: Re: [IP] Internet still reshaping history The Internet, which was the idea that there could be one single interoperable network that transcended all others, was invented by Bob Kahn and Vint Cerf. Others were involved - the idea of "internetworking" was flowering, but the timing was very clear - ithappened in the middle of the decade of the '70's. That was when theidea was reduced to practice. It is true that in 1967-68, Taylor and Licklider wrote about a "network of networks" - this was a brilliant concept, but it was not yet reducedto practice. Taylor and Licklider based their vision on the "motherof all demos" done by Doug Englebart at SRI.Indeed, the idea of information sharing goes back farther. ProjectMAC and the idea of a time-sharing system that allowed informatino sharing in a utility goes back to John McCarthy.However, the idea that the ARPANET was an "Internet" is pretty wrong. It was one of the threads creating the opportunity for the Internet.But, if you want to find the beginning of packet-based communications, why not look at Sussenguth's commercialization of IBM's SNA. Packets, of course, are not "internetworking".The real problem with these "anniversaries" is that the idea of a big "ah-ha" by one lone inventor is the kind of stupid history that gets promoted by people who are too invested in their egos. Almost everygood idea is the evolutionary result of many, ,many progenitors. On 09/06/2009 04:42 PM, David Farber wrote:Why do people say the internet is 40 years old. Is the Arpanet the Internet -- I think not. djf Begin forwarded message: From: Suzanne Johnson <fuhn () pobox com> Date: September 6, 2009 4:38:05 PM EDT To: Dave () farber net Subject: Internet still reshaping history At 40, the Internet still reshaping history by Stephen Shankland At the time, it would have been hard to predict which of these events 40 years ago would prove to be most momentous: * Humans step out of a spaceship and walk on the moon. * The Woodstock concert becomes a seminal cultural moment for the baby-boomer generation. * A New York City police raid leads to the Stonewall riots and modern gay-rights movement.* A handful of engineers at UCLA send some data from one computer toanother.You may disagree, but in my opinion, it's the last of the list: fourdecades ago today, the Internet was born. ....clip...... http://news.cnet.com/8301-30685_3-10323175-264.html?tag=rtcol;inTheNewsNow ------------------------------------------- Archives: https://www.listbox.com/member/archive/247/=now RSS Feed: https://www.listbox.com/member/archive/rss/247/ Powered by Listbox: http://www.listbox.comArchives-- Richard Bennett Research Fellow Information Technology and Innovation Foundation Washington, DC ------------------------------------------- Archives: https://www.listbox.com/member/archive/247/=now RSS Feed: https://www.listbox.com/member/archive/rss/247/ Powered by Listbox: http://www.listbox.com ----- End forwarded message ---------- End forwarded message -----Archives
-- Richard Bennett Research Fellow Information Technology and Innovation Foundation Washington, DC ------------------------------------------- Archives: https://www.listbox.com/member/archive/247/=now RSS Feed: https://www.listbox.com/member/archive/rss/247/ Powered by Listbox: http://www.listbox.com
Current thread:
- Re: Internet still reshaping history David Farber (Sep 07)
- <Possible follow-ups>
- Internet still reshaping history David Farber (Sep 07)
- Internet still reshaping history David Farber (Sep 07)
- Internet still reshaping history David Farber (Sep 07)
- Re: Internet still reshaping history David Farber (Sep 07)
- Re: Internet still reshaping history David Farber (Sep 08)
- Re: Internet still reshaping history David Farber (Sep 08)
- Re: Internet still reshaping history David Farber (Sep 08)
- Internet still reshaping history Dave Farber (Sep 08)
- Internet still reshaping history David Farber (Sep 09)
- Internet still reshaping history Dave Farber (Sep 09)
- Internet still reshaping history Dave Farber (Sep 09)
- Re: Internet still reshaping history David Farber (Sep 10)