Talk:ARPANET

From Academic Kids

This is an intermediate state. Eventually the stuff about the invention of packet switching is going to get moved to the packet switching page (just as some of this was moved here from the History of the Internet page), and more about the history of the ARPANet is going to get added here, as I have time to write it.. This is an improvement on what used to be here, though. Noel 08:07, 24 Aug 2003 (UTC)

---

Not sure where to put this (from November 21 and 1969):

The first ARPANET link is established.

--mav 05:54, 22 Nov 2003 (UTC)~

I've expanded it.

Contents

1 Data rate 2 Wesley Clark 3 images? 4 Leonard Kleinrock 5 Date clash

Data rate

I'm not sure about the nearby material that says they were connected "using modems": I think the link speeds were of the order of 56k, and the modems of that era were not even remotely near that speed (think 1200 bps): the links were more probably DS0 leased lines. -- The Anome 09:25, 6 Feb 2004 (UTC)

No, "modem" is right - that's a generic term for a device ("modulator - demodulator") that transforms a signal in one form (in this case, a bit stream) into another form (analog signals in this case). The ARPANet modems were provided by ATT; BBN's hardware only included the high-speed serial interfaces. The lines were indeed leased special high-bandwidth lines (but I don't think the concept "DS0" existed back then), again, provided by AT&T. Noel 20:24, 1 Mar 2004 (UTC)

PS: When I said "high-bandwidth", I meant it in the technical sense (i.e. high band-pass), not as a synonym for "high bit rate" (although of course with the former it's easy to provide the latter). Noel 21:10, 1 Mar 2004 (UTC)

According to [1] (http://www.2sparrows.org/Sean/rit/t1e1.htm) "The T1 standard was developed in the United States in the early 1960's." This [2] (http://www.privateline.com/TelephoneHistory4/History4.htm) says "Also in 1963 digital carrier techniques were introduced. [...] T1 quickly became the backbone of long distance toll service and then the primary handler of local transmission between central offices." So it's quite plausible that the circuit was part of a then-existing digital trunk system. However, it would be nice to have a reference which can settle this one way or another. -- The Anome 12:39, 27 Sep 2004 (UTC)

[3] (http://groups.google.com/groups?q=arpanet+data+links&start=10&hl=en&lr=&ie=UTF-8&client=firefox-a&selm=telecom12.891.1%40eecs.nwu.edu&rnum=16) describes original link speeds as having been 56k. For 1960s modems, even 600 bps over an end-to-end analog connection was "high speed". The nearest other possible solution were the FDM interleaving solutions. But why would anyone install this, when T1 was commercially available from 1962 onwards? Allowing for bit-robbing, a single DS0 demultiplexed from a T1 would be 56k, exactly the data rates quoted in the source. -- The Anome 12:57, 27 Sep 2004 (UTC)

Phil Karn also says here [4] (http://groups.google.com/groups?q=arpanet+data+links&start=20&hl=en&lr=&ie=UTF-8&client=firefox-a&selm=1990Jul24.190156.12202%40bellcore-2.bellcore.com&rnum=28) that "Actually, the ARPANET backbone links were almost all 56kb/s." -- The Anome 13:05, 27 Sep 2004 (UTC)

I simply don't know the details of how AT+T provided the bandwidth. It may well have been a breakout of a T1 line. I have looked through everything I have, and nothing says anything about it. I do know (from the BBN proposal, in response to the original ARPA ARPANET RFP) that the interface between the IMP and modem was two wires in each direction - a data line and a clock line. So I guess the bit recovery was done by the AT+T gear - maybe, because for all I know AT+T just carried those two signals through from one end to the other, mostly unchanged, and didn't do clock recovery at the far end of each link. Turning the bit stream into packets (framing - i.e. figuring out where the byte boundaries was, and then turning the byte stream into packets) was done by the BBN hardware in the IMPs. The AT+T "modems" (they didn't do everything we now think of modems as doing) were huge - at MIT (which had two AT+T lines early on, when I first became responsible for the IMPs at MIT, and a third later on) the AT+T gear was in a giant rack (about 6' high, and 2.5' square, which was mostly full of stuff). My guess would be that the bandwidth was provided in different ways, at different times and on different links, but that is just a guess. Sorry I can't help more - that's all I know and/or can find out. Noel (talk) 04:38, 11 Nov 2004 (UTC)

Wesley Clark

Well, there are a zillion links to Wesley Clark, and almost certainly all of them mean the general. So although I'd like to make it a disambig page, and then have separate pages for the general and the computer scientist, it's probably too much work to be worth it. If I can figure out the computer scientist's middle name, we can put his page under that (or. more likely, his initial), with a note somewhere on the main Wesley Clark page to refer to it. Noel 17:04, 21 Mar 2004 (UTC)

I believe the MIT Wesley Clark you wish to disambiguate is Wesley A. Clark, as referenced here: Participants in the LINC Evaluation Program (http://history.nih.gov/exhibits/linc/docs/page_10.html) -- Tracy 10:25, 27 Sep 2004 (UTC)

Oooh, thanks for dredging that up! Noel (talk) 12:00, 8 Dec 2004 (UTC)

images?

Leonard Kleinrock

I e-mailed Leonard and he outlined that the following was incorrect in the ((cur) (last) 08:23, 4 Nov 2004 Jumbuck m (robot Adding:id)) entry:

This is what was originally in the entry: "Leonard Kleinrock had performed tests on store and forward message systems in 1961, and wrote a very important book in 1964 covering queueing theory and routing in store and forward networks, although this work did not include the concept of breaking a user's message up into smaller units for transmission through the network."

Leonard wrote to me with the following:

Dear Kim,

...

This is not true. In fact, in my July 1962 publication (see my website: www.lk.cs.ucla.edu) I talk about chopping messages into fixed length units and transmitting them separately over the link. That material also appeared in my December 1962 PhD Dissertation which was later published as the book that is quoted, "Communication Nets"; you can easily find the description in that book in Section 5.3. Importantly, you should note that my July 1962 publication was months ahead of Paul Baran's and years ahead of anything that Donald Davies did. Basically, what I did for my PhD research in 1961-1962 was to establish a mathematical theory of packet networks which uncovered the underlying principles that drives today's Internet.

I would appreciate it if you would correct the entry in that link.

Best of luck. Len Kleinrock

I have updated the page accordingly. However there does seem to be some dispute - so I would like updaters to consult with Lenoard Kleirock before making changes his e-mail address is at www.lk.cs.ucla.edu - it would be even better if we could get Paul Baran and Donalad Davies in the dialog to thrash this out amongst themselves - they could then propose a nice summary that would sort out any inconsistencies - So whoever wants to take this on please be my guest! -- Kim Meyrick 21:44, 10 Nov 2004

You need to be careful with Kleinrock's claims, because he has an axe to grind. If you read Baran's page (http://www.rand.org/about/history/baran.html), which I find to be equally incorrect, Baran claims all the credit. Alas, to talk to Donald Davies, you'll need a spirit medium, because he is no longer with us. Before he died, he did dispute Kleinrock's claim. And, needless to say, if you think Baran and Kleinrock are going to agree on a division of the credit, well...

As to the 1964 book, here's what section 5.3 says:

"The server picks the next message in the queue and performs one unit of service upon it (i.e. services this message for exactly Q seconds). At the end of this time interval, the message leaves the system if its service (transmission) is finished; if not, it joins the end of the queue with its service partially completed .."

And that is the entirety of what the book says about the topic. Although no details at all are given, the most plausible reading of what was meant is that some kind of link-local fragmentation/reassembly is being used, not end-end packets of the sort described in Baran's work, and now used in the Internet.

I have the 1961 PhD proposal, and I'll look at that now. However, I suggest you be wary about rushing in where angels fear to tread! I would have to go check Hafner's book, but my vague recollection is that she got it about right. Noel (talk) 02:13, 11 Nov 2004 (UTC)

I have now checked both the 1961 thesis proposal, and the 1962 RLE QTR - the former says nothing at all, and the latter has exactly the same language as the book (quoted above). It's worth noting that the book is basically entirely about queueing theory, and is not about engineering a data communication network. The examples given in the book are principally telegram networks, although trucking networks also figure in some cases.

I also checked Hafner's book, and she doesn't explicitly rate Kleinrock's contribution, although she implicitly seems to credit Baran and Davies with a key role in coming up with the idea of packet data networks. I think there's something in that - they definitely were thinking of practical data communication systems, whereas Kleinrock's work is very much focused on the theory of queueing, which obviously has a much wider application than just packet networks. And of course Licklider and Taylor were the people who actually had the vision to build a real network. Noel (talk) 04:10, 11 Nov 2004 (UTC)

PS: DARPA's own history of the ARPANet (see article) nowhere mentions Kleinrock's work as a source of inspiration - but does prominently cite Baran and Davies. Not that I wholly believe that either (I think Roberts' quote about relying on Kleinrock's work is probably accurate), but more to show you that the picture painted by Kleinrock is not universally shared, and that you shouldn't rely on any one source. See also my recent comments at Talk:Packet switching on the subject of Kleinrock's proper share of the credit. Noel (talk) 02:44, 11 Nov 2004 (UTC)

Leonard Kleinrock has responded with the following which i post on his behalf:

"Both comments by jnc miss the point. I was not suggesting that others not be mentioned. In fact, I include the contributions of Baran and Davies in every presentation I make regarding the history of the Internet Jnc's referral to bogus claims is amazing! It may be that he has not looked at the source documents. Anyone familiar with those documents would be able to provide a more balanced view of the early work than he has articluted." -- Kim Meyrick 18:00, 12 Nov 2004

I had to laugh at "I include the contributions of Baran and Davies in every presentation I make regarding the history of the Internet". If you look at every page he has linked to from his "The History of the Internet" links page (http://www.lk.cs.ucla.edu/internet_history.html), most of which he wrote, you will not find a single instance of Baran's name, and Davies is only mentioned once, as the originator of the term "packet".

I don't know which documents he is speaking of - I have read all the ones he has posted, and a great number more besides (although I don't have access to RAND's working papers, or the internal archives of DARPA). If he has more original, contemporary documents he'd like to post, I'd be happy to review them. Noel (talk) 22:41, 12 Nov 2004 (UTC)

Date clash

Ah, the joys of people adding stuff to articles piece-meal. At ARPANET#Initial ARPANET deployment it says:

the first message ever sent over the ARPANET; it took place at 22:30PM on October 29, 1969.

and then later on it says:

The first ARPANET link was established on November 21, 1969, between an IMP at UCLA and another at SRI's

So how could a message have been sent on Oct 29 when the first link allegedly wasn't there until November 21? Alas, I don't have the time right now to ressarch this: can someone else straighten this out? Noel (talk) 00:41, 6 Jan 2005 (UTC)