From rdandi at luiss.it Wed Sep 22 10:24:12 2004 From: rdandi at luiss.it (rdandi at luiss.it) Date: Wed, 22 Sep 2004 19:24:12 +0200 Subject: [ih] questions on arpanet Message-ID: <743374743881.743881743374@luiss.it> Hi all! I am new in this list. I am Roberto Dandi, Italian PhD in organizational behavior, working at Luiss (Roma, Italy). I am collecting studies and data on the organizational and social issues of Arpanet. In particular I am interested in studying how and to what extent CMC technologies (email, online RFCs...) impacted on coordination and identification within the early arpanet community. I have three questions: 1) Email traffic: is there any statistics on the use of email in arpanet in the period 1972-1978? In some documents I found reference to a study according to which in 1973 email traffic was about 75% of the whole traffic in arpanet. Do you know where can I find this study? In RFCs I found only traffic statistics (by McKenzie) that do not distinguish email traffic. 2) Host network traffic: is there any statistics on the transfer of information from host to host? I mean, for example, how many data UCLA transferred from/to SRI or from/to other nodes? (in the period 1969-1978). Again, RFCs by McKenzie report synthetic traffic statistics, they do not specify the one to one relations. 3) The hardest question (I think): members affiliations and status. Studying RFCs I found the affiliations of the RFC authors in NWG (see attached file). I am interested also in finding which are the previous affiliations of these authors (before entering the arpanet) and the status they had during arpanet (senior or junior researcher). Does anybody know how can I fill in the blank spaces in my XL file? Thanks so much best regards Roberto Dandi -- Roberto Dandi Ph. D. in Organisational Behaviour at Universit? degli Studi del Molise (CB, Italy) c/o Scuola di Management LUISS Guido Carli Via Tommasini 1 00162 Roma, Italy tel. +39.06.86506555, fax +39.06.86506547 e-mail: rdandi at luiss.it web-site: http://www.geocities.com/robertodandi Your mail has been scanned by InterScan VirusWall. -------------- next part -------------- A non-text attachment was scrubbed... Name: Arpanet members.xls Type: application/vnd.ms-excel Size: 20480 bytes Desc: not available URL: From chris at cs.utexas.edu Wed Sep 22 11:24:08 2004 From: chris at cs.utexas.edu (Chris Edmondson-Yurkanan) Date: Wed, 22 Sep 2004 11:24:08 -0700 Subject: [ih] questions on arpanet In-Reply-To: <743374743881.743881743374@luiss.it> References: <743374743881.743881743374@luiss.it> Message-ID: <97CB45A0-0CC4-11D9-A5FC-000D932D00AE@cs.utexas.edu> Roberto, Re here's some feedback your questions: Q1 and Q2) what's the name of the study that you are referring to.... you should provide all info, such as authors, detailed date, name, does it have a NIC #, does it have a BBN #, or a Network measurement note #, etc... That will help us help you. Also, I recently found Postel's binder of all 28 Network Measurement Notes, but we haven't put any of it online yet... I don't recall whether any of them had traffic analyzed by application. You need to realize that the email applications used FTP to transfer the email messages, so it wasn't just a simple matter of looking at an Arpanet header (as we do today by examining the TCP port#s). Q3) your excel sheet * Your "status field" may need to be expanded for your research project... It appears that you want to know whether someone was a student(junior) or professor(senior) if they were academic, and otherwise they were "senior" whether they were a user, programmer, manager??? For your study, you might want their title and role, and then you can characterize the person multiple ways, sometimes using their official characterization and sometimes use the role that they took in the Arpanet. Plus, your study is over a long time period, so most juniors became seniors .... * there are 2 Crockers: Steve and Dave Good luck, Chris On Sep 22, 2004, at 10:24 AM, rdandi at luiss.it wrote: > > Hi all! > > I am new in this list. I am Roberto Dandi, Italian PhD in > organizational behavior, working at Luiss (Roma, Italy). I am > collecting studies and data on the organizational and social issues of > Arpanet. In particular I am interested in studying how and to what > extent CMC technologies (email, online RFCs...) impacted on > coordination and identification within the early arpanet community. > > I have three questions: > > 1) Email traffic: is there any statistics on the use of email in > arpanet in the period 1972-1978? In some documents I found reference > to a study according to which in 1973 email traffic was about 75% of > the whole traffic in arpanet. Do you know where can I find this study? > In RFCs I found only traffic statistics (by McKenzie) that do not > distinguish email traffic. > > 2) Host network traffic: is there any statistics on the transfer of > information from host to host? I mean, for example, how many data UCLA > transferred from/to SRI or from/to other nodes? (in the period > 1969-1978). Again, RFCs by McKenzie report synthetic traffic > statistics, they do not specify the one to one relations. > > 3) The hardest question (I think): members affiliations and status. > Studying RFCs I found the affiliations of the RFC authors in NWG (see > attached file). I am interested also in finding which are the previous > affiliations of these authors (before entering the arpanet) and the > status they had during arpanet (senior or junior researcher). Does > anybody know how can I fill in the blank spaces in my XL file? > > Thanks so much > best regards > Roberto Dandi > > -- > Roberto Dandi > Ph. D. in Organisational Behaviour > at Universit? degli Studi del Molise (CB, Italy) > c/o Scuola di Management > LUISS Guido Carli > Via Tommasini 1 > 00162 Roma, Italy > tel. +39.06.86506555, fax +39.06.86506547 > e-mail: rdandi at luiss.it > web-site: http://www.geocities.com/robertodandi > > > Your mail has been scanned by InterScan VirusWall. Chris Edmondson-Yurkanan (cey at postel.org) Postel Visiting Scholar; USC/ISI (8/16 to 12/15/2004) +1 310 448 9152 (fax +1 310 448 9300) 4676 Admiralty Way, Suite 1001 Marina del Rey, CA, 90292-6695 Permanent contact info: www.cs.utexas.edu/users/chris/ From the.map at alum.mit.edu Wed Sep 22 17:05:47 2004 From: the.map at alum.mit.edu (Mike Padlipsky) Date: Wed, 22 Sep 2004 17:05:47 -0700 Subject: [ih] questions on arpanet In-Reply-To: <97CB45A0-0CC4-11D9-A5FC-000D932D00AE@cs.utexas.edu> References: <743374743881.743881743374@luiss.it> <743374743881.743881743374@luiss.it> Message-ID: <5.2.1.1.1.20040922160109.00b13d90@mail.lafn.org> At 11:24 AM 9/22/2004, Chris Edmondson-Yurkanan wrote: >You need to realize that the email applications used FTP to transfer the >email messages, it's also important to realize that rfc's weren't distributed over the 'net until some point in time which i'll leave as an exercise to the scholarly but believe i know was after november of 1974, since that's when rfc 666 was published, 'off-line', as was the then-prevailing modality to the best of my recollection. [and as regular readers of this list will perhaps recall, it still hasn't been placed 'on-line' ... unless, of course, the well-known bottleneck has finally done so and lacked the courtesy to inform me he has; or -- just as a purist's point -- given the state of my middlemiddleaged memory, perhaps it _was_ published on-line and the bottleneck removed it, but in fact i rather doubt that.] indeed, not only weren't the rfc's 'on-line' during the formative years, but most of the key protocol design work [host-host, telnet, and of course ftp] was done at the various protocol committee/working group meetings, sometimes based on discussions we might have had with our local colleagues before the meetings but that was probably the exception, but certainly not via what was called netmail at the time, since it all took place before netmail was available [other than by telneting in to a guest account and using intra-host mail, which doesn't really count]. granted, we might well have e-discussed when and where to have the 'new' telnet meeting over the 'net, but i wouldn't vouch for it -- and i was heavily involved in the meeting. sure, it's 'only' anecdotal evidence and others' recollections might well differ, but my own is that we 'always' relied primarily on meetings for hammering out the design consensuses [or consenses, if you prefer] that were the protocols, not net- or e- mail. certainly on both versions of the host-front end protocol [the later of which was in 1984, repeat 19_8_4], and 'neted', but that might be cheating because i chaired both groups. speaking of exercises for the scholarly, b/t/w, does anybody know who's responsible for the noxiously cutesy 'e-mail' locution, and when? '73 feels awfully early to me, dr. dandi's reported referenced study to the contrary notwithstanding [though perhaps the study was a retrospective one ... and probably an inaccurate one anyway, since in '73 i suspect that at least half the traffic consisted of host-host protocol 'eco' commands, thanks to the perversity of the bbn 'ncp']. cheers, map [whose shoulder problems caused him to break down some time ago and create a 'signature' file to apologize for the lack of his formerly customary e-volubility -- and who's been employing shiftless typing for a long time now to spare his wristsnfingers, in case you didn't know ... and who's further broken down and done http://www.lafn.org/~ba213/mapstuff.html , rather grudgingly] From London at imagina.com Wed Sep 22 20:46:23 2004 From: London at imagina.com (Ralph & Bobbi London) Date: Wed, 22 Sep 2004 20:46:23 -0700 Subject: [ih] First Use of Term email Message-ID: <234E271A-0D13-11D9-82FD-0003939649A6@imagina.com> Do we know the first use of the term "email" in a written document? I'm merely curious. Ralph London From rdandi at luiss.it Thu Sep 23 08:25:35 2004 From: rdandi at luiss.it (rdandi at luiss.it) Date: Thu, 23 Sep 2004 17:25:35 +0200 Subject: [ih] questions on arpanet (2) Message-ID: <75e4fd76260e.76260e75e4fd@luiss.it> Hi all, Thanks to everybody for the fast replies. I am very excited and thankful also for this opportunity to communicate with the pioneers who were in the Arpanet community!!! The goal of my research is to find a connection between the social/organizational context of the Arpanet community and the use of email and CMC by those who actually developed these tools for the first time. I would like to get some quantitative data on these issues. Here are some further explanations of my previous questions and other questions: 1) the study that reports that in 1973 email was composing 75% of all ARPANET traffic is a ghost. Several websites dealing with internet history cite this study, but none of them shows the exact reference. They just say "an Arpa study reports"... For example: Hobbes' Internet Timeline v5.3 http://oceanpark.com/webmuseum/internet_timeline/ HistoryMole.com http://www.historymole.com/cgi-bin/main/results.pl?type=theme&theme=Internet Tom Sheldon's site http://www.linktionary.com/i/internet_history.html 2) regarding the traffic between specific IMPs.. I want to collect some network data that can show me to what extent there was an exchange of information among which nodes of Arpanet. Given that something on this exists, does anybody has some report on this traffic during the period 1969-1978? Is there something similar concerning email traffic? 3) regarding the file attached. I need to collect some attribute data on the members of the Network Working Group who co-authored RFCs during 1969-1978 (I analyzed the first 735 RFCs and got these 97 names). Through RFCs themselves I found people affilitiations. However I am interested also in the past affiliations of each member (my guess is that several members knew each other before joining Arpanet or at least had the same backgrounds as they had the same experiences - at same universities, companies, laboratories). Then I need to know the status (junior or senior) of each member in the list. If someone knows also their title (PhD student or professor, programmer or manager...) it would be great. I have to track also changes in status (so for example Steve Crocker got his PhD in 1971, so he was junior up to 1971, from then on I code him "senior", see the attached file). 4) When RFCs went online? Map Pradlipsky said after November 1974. Why so late? (Arpanet had the tools to share information online). 5) One research questions is: "Did the spread of email increase interorganizational collaboration?" Dave Crocker said yes. I am examining the co-authorship of RFCs to assess the amount of collaboration between different Arpanet research sites. However, I have to compare a pre-email introduction period with a post-email introduction period, when email was widely used. Is December 1972 a right cutoff point? I know... it's a lot of information... thanks in any case!! best Roberto Dandi -- Roberto Dandi Ph. D. in Organisational Behaviour at Universit? degli Studi del Molise (CB, Italy) c/o Scuola di Management LUISS Guido Carli Via Tommasini 1 00162 Roma, Italy tel. +39.06.86506555, fax +39.06.86506547 e-mail: rdandi at luiss.it web-site: http://www.geocities.com/robertodandi ----- Original Message ----- From: Dave Crocker Date: Thursday, September 23, 2004 2:54 am Subject: Re: [ih] questions on arpanet > *********************** > Your mail has been scanned by InterScan VirusWall. > ***********-*********** > > > On Wed, 22 Sep 2004 19:24:12 +0200, rdandi at luiss.it wrote: > >particular I am interested in studying how and to what extent CMC > technologies (email, online RFCs...) impacted on coordination and > identification within the early arpanet community. > > the impact was tremendous. it quickly became key to multi- > organization > collaboration. > > > > 1) Email traffic: is there any statistics on the use of email > in > arpanet in the period 1972-1978? In some documents I found > reference to > a study according to which in 1973 email traffic was about 75% of > the > whole traffic in arpanet. Do you know where can I find this study? > In > RFCs I found only traffic statistics (by McKenzie) that do not > distinguish email traffic. > > I do not have numbers, but the reports I remember hearing through > the > 70's and even 80's was that Telnet, not email, dominated network > traffic. > > > d/ > > > Your mail has been scanned by InterScan VirusWall. -------------- next part -------------- A non-text attachment was scrubbed... Name: arpanet members2.xls Type: application/octet-stream Size: 22016 bytes Desc: not available URL: From craig at bbn.com Wed Sep 22 11:39:03 2004 From: craig at bbn.com (Craig Partridge) Date: Wed, 22 Sep 2004 14:39:03 -0400 (EDT) Subject: [ih] questions on arpane Message-ID: <20040922183903.DB191CA3A@albino.ir.bbn.com> In message <743374743881.743881743374 at luiss.it>, rdandi at luiss.it writes: >1) Email traffic: is there any statistics on the use of email in arpanet in= > the period 1972-1978? In some documents I found reference to a study= > according to which in 1973 email traffic was about 75% of the whole= > traffic in arpanet. Do you know where can I find this study? In RFCs I= > found only traffic statistics (by McKenzie) that do not distinguish email= > traffic.=0D I believe that number is right, but I don't recall how it was collected. >2) Host network traffic: is there any statistics on the transfer of= > information from host to host? I mean, for example, how many data UCLA= > transferred from/to SRI or from/to other nodes? (in the period 1969-1978).= > Again, RFCs by McKenzie report synthetic traffic statistics, they do not= > specify the one to one relations.=0D Those numbers were reported weekly, as I recall, by BBN's network operations center right through the 1980s. The reports tended to be by IMP port (i.e. connection to an IMP) and what I remember seeing most of the time was simply per-port traffic counts, but there were also IMP-to-IMP traffic counts. We knew, for instance, that traffic from CSNET Relay's IMP to SEISMO's IMP (the UUCP gateway) was very heavy. [These reports where used in DNS debugging to realize that people weren't updating their root files -- it was discovered that ISI-C's old IMP port was getting blasted with traffic long after the machine (which had been a root server) was disconnected] The reports were sent via email. I have no clue if anyone kept them (Dave Mills perhaps?). I suspect they also were part of the monthly status reports to DARPA -- I don't know where those status reports are (I'd assume in the US National Archives by now). >3) The hardest question (I think): members affiliations and status.= > Studying RFCs I found the affiliations of the RFC authors in NWG (see= > attached file). I am interested also in finding which are the previous= > affiliations of these authors (before entering the arpanet) and the status= > they had during arpanet (senior or junior researcher). Does anybody know= > how can I fill in the blank spaces in my XL file? I suspect this is actually the easiest -- send a list of names to this list and I'll bet we can collectively fill it in (certainly status, in some cases with titles). It is still a small world :-) Thanks! Craig Partridge Chief Scientist, BBN Technologies From dhc2 at dcrocker.net Wed Sep 22 17:54:26 2004 From: dhc2 at dcrocker.net (Dave Crocker) Date: Wed, 22 Sep 2004 17:54:26 -0700 Subject: [ih] questions on arpanet In-Reply-To: <743374743881.743881743374@luiss.it> Message-ID: <2004922175426.536369@bbprime> On Wed, 22 Sep 2004 19:24:12 +0200, rdandi at luiss.it wrote: >particular I am interested in studying how and to what extent CMC technologies (email, online RFCs...) impacted on coordination and identification within the early arpanet community. the impact was tremendous. it quickly became key to multi-organization collaboration. > 1) Email traffic: is there any statistics on the use of email in arpanet in the period 1972-1978? In some documents I found reference to a study according to which in 1973 email traffic was about 75% of the whole traffic in arpanet. Do you know where can I find this study? In RFCs I found only traffic statistics (by McKenzie) that do not distinguish email traffic. I do not have numbers, but the reports I remember hearing through the 70's and even 80's was that Telnet, not email, dominated network traffic. d/ From braden at ISI.EDU Thu Sep 23 13:24:30 2004 From: braden at ISI.EDU (Bob Braden) Date: Thu, 23 Sep 2004 13:24:30 -0700 (PDT) Subject: [ih] Re: internet-history Digest, Vol 2, Issue 4 Message-ID: <200409232024.NAA21029@gra.isi.edu> *> *> I do not have numbers, but the reports I remember hearing through the *> 70's and even 80's was that Telnet, not email, dominated network *> traffic. *> *> *> d/ Perhaps by message count, but not by byte count? Telnet in those days was entirely echoplex char/cha. The dominating Telnet traffic was probably dominated by remote users to the Tenex/TOPS20 farms at ISI and BBN. Bob From braden at ISI.EDU Thu Sep 23 13:27:16 2004 From: braden at ISI.EDU (Bob Braden) Date: Thu, 23 Sep 2004 13:27:16 -0700 (PDT) Subject: [ih] Re: internet-history Digest, Vol 2, Issue 4 Message-ID: <200409232027.NAA21032@gra.isi.edu> I would like to make a comment about this discussion of ARPAnet history. We pioneers who were there at the time have at best hazy memories of what happened when. The only reliable record from those days is the RFC series. If it was written down, it is still preserved. What was not written down was largely lost. Twenty years from now, Internet historians will be asking about what happened to the Internet in the 1990s and early 21st century. (Where) is it being written down? Bob Braden From chris at cs.utexas.edu Thu Sep 23 15:52:51 2004 From: chris at cs.utexas.edu (Chris Edmondson-Yurkanan) Date: Thu, 23 Sep 2004 15:52:51 -0700 Subject: [ih] Re: internet-history Digest, Vol 2, Issue 4 In-Reply-To: <200409232027.NAA21032@gra.isi.edu> References: <200409232027.NAA21032@gra.isi.edu> Message-ID: <4C982CF6-0DB3-11D9-A5FC-000D932D00AE@cs.utexas.edu> I have observed that the network research community from the 70s also created their own smaller sets of series to supplement the RFCs. Unfortunately, most of these series are not online (possibly because they didn't have an RFC-editor like role?). However, those documents are mostly available to the researcher who physically visits the archives in whatever university archive (or museum) it happens to reside in. (See online status at the bottom of this mail) For a short list of document series from the 70s besides the RFCs see the list at the bottom of this mail. Just as the IETF creates working groups, so did the ARPAnet and Internet researchers. I assume that the phenomenon is very natural (fractal-like?): that once a small community project grows to a certain size, that researchers like to create additional series that are distributed to just their small working group... with occasional publicizing in the larger community (e.g. RFC619 Network Measurement Note 18, or RFC 569 Network Using Note 1). What's interesting to me is that today's IETF working groups use internet-drafts, mail archives, and "additional web pages" to supplement any RFCs that the working group generates. 1) Unfortunately, the links to the mail archives and web pages of recently concluded working groups are not all archived on the ietf server; and thus even an attempt to keep them has failed. (Plus, we already know that the internet-drafts are not archived... no need to discuss this here....) 2) The concept of an ordered (numbered) series is not used by the IETF working groups, except for the version # of a protocol internet-draft in development. University Tech Reports do use the numbered series...and good news here: the CS tech reports are archived by the community. So, to expand on Bob's point: * the network research community has to at least find a mechanism and archival home for design discussions that are not in tech reports or workshop/conference proceedings (such as old ietf working group mail and internet-drafts). * I wonder if the "numbered series" would even work anymore... (I assume that it could work in an individual working group). The beauty of a numbered series, is that some "editor" decided that it was relevant to the working group and gave it an archival #, date, and location for others to access, even if it was also in a tech report or conference paper. * I think that a network research professional society web server could be an ideal home for all for online materials, but that would need to be ratified/supported by the entire "international" community... thanks, Chris PS: here's a list of just some of the 70's small series that supplemented the RFCs (this list does not include the network research outside of Arpanet and Internet specific protocols): IENs Internet Experiment Notes, 1977-78 design of TCP (then TCP/IP) INWGs (whose official title started off as International Packet Network Working Group)...they had experiment series, general series, etc) and specific project documents, such as the: "M.I.T. Lab. for Computer Science Network Implementation Notes" (i.e. a "tech report"?) And then there are the working groups that created their own series.... Note, that the following were in Postel's papers, so that's why the community has access to these: Network Measurement Notes 1-28 (1972-1974, then disbanded) Network Graphics Group Notes (1-10); 1973-74 User Interest Group (USING) Notes 5-8, 11, 13; 1973-74 USERS working group Notes 1-2. 1973? PS2: The IENs will hopefully soon be online, and some INWGs are online. I expect to put some of the others online from the Postel Archives. On Sep 23, 2004, at 1:27 PM, Bob Braden wrote: > > I would like to make a comment about this discussion of ARPAnet > history. > We pioneers who were there at the time have at best hazy memories of > what happened when. The only reliable record from those days is > the RFC series. If it was written down, it is still preserved. > What was not written down was largely lost. > > Twenty years from now, Internet historians will be asking about > what happened to the Internet in the 1990s and early 21st century. > (Where) is it being written down? > > Bob Braden > Chris Edmondson-Yurkanan (cey at postel.org) Postel Visiting Scholar; USC/ISI (8/16 to 12/15/2004) +1 310 448 9152 (fax +1 310 448 9300) 4676 Admiralty Way, Suite 1001 Marina del Rey, CA, 90292-6695 Permanent contact info: www.cs.utexas.edu/users/chris/ From dhc2 at dcrocker.net Thu Sep 23 16:02:06 2004 From: dhc2 at dcrocker.net (Dave Crocker) Date: Thu, 23 Sep 2004 16:02:06 -0700 Subject: [ih] First Use of Term email In-Reply-To: <234E271A-0D13-11D9-82FD-0003939649A6@imagina.com> Message-ID: <20049231626.776601@bbprime> On Wed, 22 Sep 2004 20:46:23 -0700, Ralph & Bobbi London wrote: > Do we know the first use of the term "email" in a written document? > I'm merely curious. I am pretty sure we were using the term by 1975, but perhaps other folks have different memories. I certainly cannot remember when the term was first used. d/ From faber at ISI.EDU Fri Sep 24 10:27:34 2004 From: faber at ISI.EDU (Ted Faber) Date: Fri, 24 Sep 2004 10:27:34 -0700 Subject: [ih] Re: internet-history Digest, Vol 2, Issue 4 In-Reply-To: References: <200409232027.NAA21032@gra.isi.edu> Message-ID: <20040924172734.GC38136@pun.isi.edu> On Fri, Sep 24, 2004 at 01:56:28AM +0100, Lloyd Wood wrote: > On Thu, 23 Sep 2004, Bob Braden wrote: > > > I would like to make a comment about this discussion of ARPAnet history. > > We pioneers who were there at the time have at best hazy memories of > > what happened when. The only reliable record from those days is > > the RFC series. If it was written down, it is still preserved. > > What was not written down was largely lost. > > ...which is rather convenient for those involved, in being able to > claim that what was written down was the result of Really Careful > Thought and Much Discussion of Alternatives, alas and most oddly not > visible anywhere. Sadly that's often true, but the best of these sorts of documents do document the paths not taken and reasons why the authors chose one over another. > I'm rather worried that academics will be taking all the credit, > simply because academics are self-documenting Dr Johnsons. Always > scribble, scribble, scribble, eh? And I'm always being told I don't write enough... Seriously, there are a variety of places for researchers and engineers from all over to get their ideas published these days. From white papers to traditional conferences and workshops to google-archived web pages, it's easier than ever to get an idea on the public record. Furthermore, because of patent concerns, non-academics have financial incentives to publicize their ideas and innovations to stake out turf. Academics are certainly encouraged to take credit, but non-academics are in a better position than ever to stake a claim on new ideas, especially in the networking field. -- Ted Faber http://www.isi.edu/~faber PGP: http://www.isi.edu/~faber/pubkeys.asc Unexpected attachment on this mail? See http://www.isi.edu/~faber/FAQ.html#SIG -------------- next part -------------- A non-text attachment was scrubbed... Name: not available Type: application/pgp-signature Size: 187 bytes Desc: not available URL: From falk at ISI.EDU Fri Sep 24 14:07:27 2004 From: falk at ISI.EDU (Aaron Falk) Date: Fri, 24 Sep 2004 14:07:27 -0700 Subject: [ih] Re: internet-history Digest, Vol 2, Issue 4 In-Reply-To: <20040924172734.GC38136@pun.isi.edu> References: <200409232027.NAA21032@gra.isi.edu> <20040924172734.GC38136@pun.isi.edu> Message-ID: On Sep 24, 2004, at 10:27 AM, Ted Faber wrote: > Seriously, there are a variety of places for researchers and engineers > from all over to get their ideas published these days. From white > papers to traditional conferences and workshops to google-archived web > pages, it's easier than ever to get an idea on the public record. > Ah, but published != archived. Where will the public record of these ideas be 30 years from now? --aaron From faber at ISI.EDU Fri Sep 24 15:20:21 2004 From: faber at ISI.EDU (Ted Faber) Date: Fri, 24 Sep 2004 15:20:21 -0700 Subject: [ih] Re: internet-history Digest, Vol 2, Issue 4 In-Reply-To: References: <200409232027.NAA21032@gra.isi.edu> <20040924172734.GC38136@pun.isi.edu> Message-ID: <20040924222021.GC60536@pun.isi.edu> On Fri, Sep 24, 2004 at 02:07:27PM -0700, Aaron Falk wrote: > > On Sep 24, 2004, at 10:27 AM, Ted Faber wrote: > > >Seriously, there are a variety of places for researchers and engineers > >from all over to get their ideas published these days. From white > >papers to traditional conferences and workshops to google-archived web > >pages, it's easier than ever to get an idea on the public record. > > > > Ah, but published != archived. Where will the public record of these > ideas be 30 years from now? Publication alone does not create archives, but wider publishing and better indexing has to make an archivist's job easier. (Please make sure one reads both sides of that "and" before sending me mail about how more information is not necessarily easier to archive.) If it's easier to do and the benefits are more accesiible to others, more people will do it. As evidence consider these: http://epguides.com/DukesofHazzard/guide.shtml http://www.tvtome.com/tvtome/servlet/EpisodeGuideSummary/showid-684/The_Dukes_of_Hazzard/ http://www.dukefarm.com/episode1.html I also suspect that no one concerned with this topic hypothesized such an archive would be created or maintained. It's fairly easy to make the archives and make them acccessible, so they happen. Existing archives also become more accesible: http://www.ced.berkeley.edu/cedarchives/ Lloyd expressed a concern that academics would take all the credit for network ideas, presumably because academics would document their work more often and also manage the archives. I'm saying that there are more opportunities to make your work known these days and that there are more archivists these days (tecnnology's making it easier). I hypothesise that that combination implies there is more chance that non-academic work will make it into a[n] historical record/archive/greatest hits collection. Immortality is easier than you thought. "If you would not be forgotten, As soon as you are dead and rotten, Either write things worthy reading, Or do things worth the writing." -- Benjamin Franklin, Poor Richard's Almanac, May 1738 -- Ted Faber http://www.isi.edu/~faber PGP: http://www.isi.edu/~faber/pubkeys.asc Unexpected attachment on this mail? See http://www.isi.edu/~faber/FAQ.html#SIG -------------- next part -------------- A non-text attachment was scrubbed... Name: not available Type: application/pgp-signature Size: 187 bytes Desc: not available URL: From touch at ISI.EDU Mon Sep 27 09:11:53 2004 From: touch at ISI.EDU (Joe Touch) Date: Mon, 27 Sep 2004 09:11:53 -0700 Subject: [ih] Re: internet-history Digest, Vol 2, Issue 4 In-Reply-To: References: <200409232027.NAA21032@gra.isi.edu> Message-ID: <41583BC9.8030502@isi.edu> Lloyd Wood wrote: ... > I'm rather worried that academics will be taking all the credit, > simply because academics are self-documenting Dr Johnsons. Always > scribble, scribble, scribble, eh? Academic institutions have concepts of time that are more than 6 months out, and endowments (translation: retirement accounts) that support this. Industry used to scribble too - tech reports, etc. But they disappear, notably when Company A is bought by Company B. Fujitsu had some nice docs about a neat architecture called the F8; by early 1990s it had disappeared except in personal copies, since the purchasing company didn't consider a library a "financial asset". History, for better or worse, belongs to those who write the books. As Ted implied, if we care about it, we need to write things down more. Don't blame academics, who participated heavily in the development of the Internet anyway, for having the time for something industry doesn't see as having a 5-minute payback. Joe -------------- next part -------------- A non-text attachment was scrubbed... Name: signature.asc Type: application/pgp-signature Size: 254 bytes Desc: OpenPGP digital signature URL: From jnc at mercury.lcs.mit.edu Mon Sep 27 11:36:52 2004 From: jnc at mercury.lcs.mit.edu (Noel Chiappa) Date: Mon, 27 Sep 2004 14:36:52 -0400 (EDT) Subject: [ih] Re: internet-history Digest, Vol 2, Issue 4 Message-ID: <20040927183652.DCBB486B02@mercury.lcs.mit.edu> > From: Lloyd Wood >> academics, who participated heavily in the development of the Internet >> anyway, > ...as their written records attest... Ahem. You may be making a joke, but I can't tell. If so, please excuse my lack of humour about this topic. Anyway, if you were serious, then some of the pople on this list WERE THERE, and although our memories may not be perfect, it's pretty simple to demonstrate the makeup of the early Internet Working Group during the formative years (76-78 or so). As I run down a mental list of the people *I personally* remember being around in '77-'78 (Vint Cerf, Bob Kahn, David Clark, David Reed, Jon Postel, Bob Braden, Danny Cohen, Ray McFarland, Jim Forgie, Jim Mathis, Jack Haverty, Bill Plummer, Ginnny Strazisar, Radia Perlman, David Mills, Ed Cain, Peter Kirstein, Andrew Hinchley, John Shoch - and apologies to those whom I've left out), and of all of them only *one* (Shoch) was not either i) an academic, or ii) someone at a DARPA-funded contractor (such as the people from SRI and BBN). If you could jog my memory by providing the names of any commercially-funded people involved with the project in this time period, that would be useful. Noel From arussell at jhu.edu Mon Sep 27 13:48:08 2004 From: arussell at jhu.edu (Andrew Russell) Date: Mon, 27 Sep 2004 16:48:08 -0400 Subject: [ih] Re: internet-history Digest, Vol 2, Issue 4 In-Reply-To: References: <200409232027.NAA21032@gra.isi.edu> <41583BC9.8030502@isi.edu> Message-ID: <89FD5C9C-10C6-11D9-B923-0050E4966508@jhu.edu> Lloyd (and all) - Given this discussion, you might enjoy a chapter by Amy Slaton and Janet Abbate called "The Hidden Lives of Standards: Technical Prescriptions and the Transformation of Work in America," in the book "Technologies of Power: Essays in Honor of Thomas Parke Hughes and Agatha Chipley Hughes" (eds. Michael Thad Allen and Gabrielle Hecht, The MIT Press, 2001). The authors argue, "To see the full range of labor implications of standards, it is necessary to look at the actual processes involved in putting standards to use. By uncovering the labor required of the people who implement, test, and deploy standardized products, this study demonstrates how standards can, in some instances, become a means of intensifying white-collar labor and an opportunity for self-made experts to assert authority." (I can't find this article online, but you might find that it's worth a trip to the library). They look at the 1983 TCP transition and the growth of Interop (among other things) to illustrate their arguments. The general point is that specification of standards is a vitally important part of the story, but it's only the beginning. Implementation (by people who, unfortunately, as Joe points out, aren't as fastidious in keeping and maintaining records) is another important part. Well-trained historians (like Slaton and Abbate) know to look for additional evidence (archived or otherwise) that might show who else was involved beyond what is recorded in RFCs, etc. Internet history isn't simply "who invented what." For example, Urs von Burg's book "The Triumph of Ethernet" (Stanford UP, 2001) shows that the history of Ethernet doesn't stop with a biography of Robert Metcalfe. Another good study that argues this general point is Greg Downey's book "Telegraph Messenger Boys" (Routledge, 2002). He also wrote an excellent article (particularly relevant to this topic) called "Virtual Webs, Physical Technologies, and Hidden Workers: The Spaces of Labor in Information Networks" in the journal "Technology and Culture" (Vol 42 No 2 April 2001, pp 209-236 - at http://muse.jhu.edu/journals/technology_and_culture/toc/tech42.2.html, but alas you'll need an individual or institutional subscription). As for the general concern (expressed by Joe, Bob, and others) about disappearing archival sources on the industry side, David Kirsch at the University of Maryland is working hard to try to preserve records and memories of failed dot-com firms. See http://businessplanarchive.org and http://www.dotcomarchive.org. If archives tend to disappear when Company A is bought by Company B, they *definitely* disappear when Company A's assets become the property of bankruptcy trustees. Andy Russell History of Science and Technology Johns Hopkins University On Sep 27, 2004, at 1:25 PM, Lloyd Wood wrote: > On Mon, 27 Sep 2004, Joe Touch wrote: > >>> I'm rather worried that academics will be taking all the credit, >>> simply because academics are self-documenting Dr Johnsons. Always >>> scribble, scribble, scribble, eh? >> >> Academic institutions have concepts of time that are more than 6 >> months >> out, and endowments (translation: retirement accounts) that support >> this. >> >> Industry used to scribble too - tech reports, etc. But they disappear, >> notably when Company A is bought by Company B. Fujitsu had some nice >> docs about a neat architecture called the F8; by early 1990s it had >> disappeared except in personal copies, since the purchasing company >> didn't consider a library a "financial asset". >> >> History, for better or worse, belongs to those who write the books. As >> Ted implied, if we care about it, we need to write things down more. >> Don't blame academics, who participated heavily in the development of >> the Internet anyway, > > ...as their written records attest... > > > for having the time for something industry doesn't >> see as having a 5-minute payback. >> >> Joe >> > > > From dpreed at reed.com Mon Sep 27 13:54:27 2004 From: dpreed at reed.com (David P. Reed) Date: Mon, 27 Sep 2004 16:54:27 -0400 Subject: [ih] Re: internet-history Digest, Vol 2, Issue 4 In-Reply-To: References: <200409232027.NAA21032@gra.isi.edu> <41583BC9.8030502@isi.edu> Message-ID: <41587E03.5060209@reed.com> I'm not sure what purpose this attempt to put people in boxes (academic/industrial) serves. Especially when the mean-spirited jerks seem to want to make something negative of that distinction. Once upon a time, we were all students, which makes us all academics in one sense. Take me for example.... but many of those who helped create the Internet Protocols have remarkably similar experiences. I have been an academic and corporate vice president and an industrial researcher in roughly equal parts during my 35 year career so far. I've been the chief designer and manager of teams that have shipped several billions of dollars worth of software products, during the days when the software industry was only 10's of billions of dollars annually world wide, and of course I've also participated in others' teams. I've written a few important papers, and have a few patents. I was one of the many who helped make the Internet protocol design what it is, and I'm proud of what *we* did, and I have a sense of what difference my participation made. So am I an academic or a commercial person? Why does this matter? Why should it? -------------- next part -------------- A non-text attachment was scrubbed... Name: dpreed.vcf Type: text/x-vcard Size: 113 bytes Desc: not available URL: From touch at ISI.EDU Mon Sep 27 15:04:01 2004 From: touch at ISI.EDU (Joe Touch) Date: Mon, 27 Sep 2004 15:04:01 -0700 Subject: [Fwd: Re: [ih] Re: internet-history Digest, Vol 2, Issue 4] Message-ID: <41588E51.9010408@isi.edu> Forwarded to the list for archive. -------------- next part -------------- An embedded message was scrubbed... From: Lloyd Wood Subject: Re: [ih] Re: internet-history Digest, Vol 2, Issue 4 Date: Mon, 27 Sep 2004 18:25:12 +0100 (BST) Size: 2839 URL: -------------- next part -------------- A non-text attachment was scrubbed... Name: signature.asc Type: application/pgp-signature Size: 254 bytes Desc: OpenPGP digital signature URL: From touch at ISI.EDU Mon Sep 27 15:03:47 2004 From: touch at ISI.EDU (Joe Touch) Date: Mon, 27 Sep 2004 15:03:47 -0700 Subject: [Fwd: Re: [ih] Re: internet-history Digest, Vol 2, Issue 4] Message-ID: <41588E43.9070904@isi.edu> I'm posting Lloyd's posts to the list again, so they show up in the archive. I'm not sure why his posts are being treated differently, and am looking into it further now. Joe -------------- next part -------------- An embedded message was scrubbed... From: Lloyd Wood Subject: Re: [ih] Re: internet-history Digest, Vol 2, Issue 4 Date: Fri, 24 Sep 2004 01:56:28 +0100 (BST) Size: 3567 URL: -------------- next part -------------- A non-text attachment was scrubbed... Name: signature.asc Type: application/pgp-signature Size: 254 bytes Desc: OpenPGP digital signature URL: From touch at ISI.EDU Mon Sep 27 15:07:45 2004 From: touch at ISI.EDU (Joe Touch) Date: Mon, 27 Sep 2004 15:07:45 -0700 Subject: [Fwd: Re: [ih] Re: internet-history Digest, Vol 2, Issue 4] Message-ID: <41588F31.9010307@isi.edu> Forwarded for the archive. -------- Original Message -------- Subject: Re: [ih] Re: internet-history Digest, Vol 2, Issue 4 Date: Fri, 24 Sep 2004 01:56:28 +0100 (BST) From: Lloyd Wood Organization: speaking for none To: Bob Braden CC: internet-history at postel.org References: <200409232027.NAA21032 at gra.isi.edu> On Thu, 23 Sep 2004, Bob Braden wrote: > I would like to make a comment about this discussion of ARPAnet history. > We pioneers who were there at the time have at best hazy memories of > what happened when. The only reliable record from those days is > the RFC series. If it was written down, it is still preserved. > What was not written down was largely lost. ...which is rather convenient for those involved, in being able to claim that what was written down was the result of Really Careful Thought and Much Discussion of Alternatives, alas and most oddly not visible anywhere. > Twenty years from now, Internet historians will be asking about > what happened to the Internet in the 1990s and early 21st century. > (Where) is it being written down? in mailing list archives, which the ietf should finally be getting its act together about hosting (and hopefully archiving) rather too late. Oh, it hasn't?, because history is still neglected? Ah, well. the end2end archives go back quite some way. I'm rather worried that academics will be taking all the credit, simply because academics are self-documenting Dr Johnsons. Always scribble, scribble, scribble, eh? L. -------------- next part -------------- A non-text attachment was scrubbed... Name: signature.asc Type: application/pgp-signature Size: 254 bytes Desc: OpenPGP digital signature URL: From touch at ISI.EDU Mon Sep 27 15:07:57 2004 From: touch at ISI.EDU (Joe Touch) Date: Mon, 27 Sep 2004 15:07:57 -0700 Subject: [Fwd: Re: [ih] Re: internet-history Digest, Vol 2, Issue 4] Message-ID: <41588F3D.7020406@isi.edu> forwarded for the archive -------- Original Message -------- Subject: Re: [ih] Re: internet-history Digest, Vol 2, Issue 4 Date: Mon, 27 Sep 2004 18:25:12 +0100 (BST) From: Lloyd Wood Organization: speaking for none To: Joe Touch CC: internet-history at postel.org, Bob Braden References: <200409232027.NAA21032 at gra.isi.edu> <41583BC9.8030502 at isi.edu> On Mon, 27 Sep 2004, Joe Touch wrote: > > I'm rather worried that academics will be taking all the credit, > > simply because academics are self-documenting Dr Johnsons. Always > > scribble, scribble, scribble, eh? > > Academic institutions have concepts of time that are more than 6 months > out, and endowments (translation: retirement accounts) that support this. > > Industry used to scribble too - tech reports, etc. But they disappear, > notably when Company A is bought by Company B. Fujitsu had some nice > docs about a neat architecture called the F8; by early 1990s it had > disappeared except in personal copies, since the purchasing company > didn't consider a library a "financial asset". > > History, for better or worse, belongs to those who write the books. As > Ted implied, if we care about it, we need to write things down more. > Don't blame academics, who participated heavily in the development of > the Internet anyway, ...as their written records attest... for having the time for something industry doesn't > see as having a 5-minute payback. > > Joe > -------------- next part -------------- A non-text attachment was scrubbed... Name: signature.asc Type: application/pgp-signature Size: 254 bytes Desc: OpenPGP digital signature URL: From touch at ISI.EDU Mon Sep 27 15:09:16 2004 From: touch at ISI.EDU (Joe Touch) Date: Mon, 27 Sep 2004 15:09:16 -0700 Subject: [ih] failed archiving of Lloyd's posts Message-ID: <41588F8C.3040004@isi.edu> Hi, all, It has been brought to my attention that the IH archives do not include Lloyd's two recent posts. I've reposted them to add to the archives. I'm not yet sure why his posts are being treated differently, and am looking into this glitch now. Joe -------------- next part -------------- A non-text attachment was scrubbed... Name: signature.asc Type: application/pgp-signature Size: 254 bytes Desc: OpenPGP digital signature URL: From touch at ISI.EDU Mon Sep 27 15:14:03 2004 From: touch at ISI.EDU (Joe Touch) Date: Mon, 27 Sep 2004 15:14:03 -0700 Subject: [ih] Re: failed archiving of Lloyd's posts In-Reply-To: <41588F8C.3040004@isi.edu> References: <41588F8C.3040004@isi.edu> Message-ID: <415890AB.2020801@isi.edu> Hi, all, It turns out that Lloyd had set X-no-archive on his emails, so this was not a failure of the list software. Joe Joe Touch wrote: > Hi, all, > > It has been brought to my attention that the IH archives do not include > Lloyd's two recent posts. I've reposted them to add to the archives. > > I'm not yet sure why his posts are being treated differently, and am > looking into this glitch now. > > Joe -------------- next part -------------- A non-text attachment was scrubbed... Name: signature.asc Type: application/pgp-signature Size: 254 bytes Desc: OpenPGP digital signature URL: From touch at ISI.EDU Mon Sep 27 15:58:12 2004 From: touch at ISI.EDU (Joe Touch) Date: Mon, 27 Sep 2004 15:58:12 -0700 Subject: [ih] Re: failed archiving of Lloyd's posts In-Reply-To: <415890AB.2020801@isi.edu> References: <41588F8C.3040004@isi.edu> <415890AB.2020801@isi.edu> Message-ID: <41589B04.6030601@isi.edu> FYI: I forwarded posts to this lists since the list has had an archive since its inception; posting to the list (IMO) constitutes permission to archive. As a result, posts with 'x-no-archive' set will be bounced from now on. I regret that some may choose not to participate as a result, but since this list is all about history, it seems nonsensical to support posts that are not archived. Joe Joe Touch wrote: > Hi, all, > > It turns out that Lloyd had set X-no-archive on his emails, so this was > not a failure of the list software. > > Joe > > Joe Touch wrote: > >> Hi, all, >> >> It has been brought to my attention that the IH archives do not >> include Lloyd's two recent posts. I've reposted them to add to the >> archives. >> >> I'm not yet sure why his posts are being treated differently, and am >> looking into this glitch now. >> >> Joe -------------- next part -------------- A non-text attachment was scrubbed... Name: signature.asc Type: application/pgp-signature Size: 254 bytes Desc: OpenPGP digital signature URL: From feinler at earthlink.net Mon Sep 27 21:14:03 2004 From: feinler at earthlink.net (Jake Feinler) Date: Mon, 27 Sep 2004 21:14:03 -0700 Subject: [ih] Writing things down - then and now Message-ID: In my lab days we were required to keep lab notebooks of every experiment we did. These were kept and were proof of work done by the individual and/or organization (academic or commercial) in the event of a controversy. The problem nowadays isn't that things aren't written down - more is written down than ever before. Unfortunately, it is mostly written as email and most organizations treat email as ephemera. Perhaps this needs to be rethought. As head of the Arpanet/DDN Network Information Center (NIC) for many years, I tried to keep the email for some of the important working groups of the time, such as TCP/IP (and was usually told I was off my rocker.) Now I wish I had kept more detail so that we wouldn't need to be having this discussion, and there would be a repository where such important "lab notebooks" could be accessed. Besides some of the discussions were not only technical but they were very amusing when discussions heated up. (Who could forget such gems as "Okay it's come to this..."Your mother wears combat boots!") I have donated all my NIC papers to the Computer History Museum in Sunnyvale, CA. In the collection I have found email pertaining to various working groups. If anything looks interesting or reasonably complete, I am sure the Museum would want to share with the public (although I have to go on record as saying I do not speak officially on their behalf, as I am a volunteer not an employee.) Regards, Jake Feinler From feinler at earthlink.net Tue Sep 28 12:50:07 2004 From: feinler at earthlink.net (Jake Feinler) Date: Tue, 28 Sep 2004 12:50:07 -0700 Subject: [ih] Location of Computer Museum Message-ID: <999934A6-1187-11D9-8EC3-003065B94CBE@earthlink.net> In my last message I mistakenly located the Computer History Museum in Sunnyvale. WRONG! (I was a little groggy at the time.) It is located in Mountain View on Shoreline Blvd. across from all the movies on the same street. For those not familiar with the museum, it has a website, www.computerhistory.org and we have docent led tours Weds., Fri, and Sat. Check out our website or, If you are in the area, please stop by; it is quite exciting! And DON'T THROW ANYTHING AWAY until you check with organizations like ours that might find gems in your desktop detritus. My apologies for the misinformation. Jake From ian.peter at ianpeter.com Thu Sep 30 23:36:11 2004 From: ian.peter at ianpeter.com (Ian Peter) Date: Fri, 1 Oct 2004 16:36:11 +1000 Subject: [ih] (LONG) - Separating TCP and IP In-Reply-To: <200409291900.i8TJ05J05010@boreas.isi.edu> Message-ID: Some of you will be aware of the Internet history site I am running at www.nethistory.info. It's a portal, and essentially aimed at less technical people. But one of the things that comes up continually is suitable archives and more in depth material, so it was interesting to see the recent discussion on this subject. Meanwhile some people here will be interested in the following email exchange with Vint Cerf, Bob Frankston and David Reed, on the subject of early tcp and ip separation. I'm reproducing it here with the permission of the participants. I'm sure some people here will want to add comments,and I am sure the record of conversation will be of interest to researchers. It's fairly long. I've edited the content a bit to get to the original discussion sequence. Enjoy! Ian Peter (Bob Frankston) While I am biased, I notice that the histories rarely mention David (Reed of course -- there are too many Davids though David Reed is not exactly unique either) despite his role in the end-to-end concept and the related separation TCP from IP. You can ask the other participants in your list for their comments on this. A number of the key concepts could be found in the design of the Ethernet though they arose more from imitating the AlohaNet than because of an explicit end-to-end design point. What makes it interesting is that it the idea of putting a reliable transport didn't even come up. I remember sitting in class when Bob (Metcalfe - too many Bobs?) spoke about his project. Of course the computers, not the network, would take care of assuring that the message got through. The insight was in making this an explicit design point of the Internet since the big Arpanet had separate computers - the IMPs - that could perform all sorts of magic including assuring a reliable transport. The defining concept of the Internet is the / between TCP and IP. Writing it as TCP/IP is unfortunate since they are very far apart. UDP/IP might be more appropriate though I just talk about TCP or IP depending on which is appropriate. (Vint Cerf) Hi Bob As I recall, David was certainly a proponent of the end/end philosophy that drove TCP. He also thought that we should use random sequence number (64 bit?) rather than the Initial Sequence Number mechanism based on clocks - I was not in favor of the random method but it had the advantage that you didn't need a quiet time. I was uncomfortable with the potential for ISN collision though. I don't recall David's role in arguing for the split of IP from TCP - what I remember most vividly is Danny Cohen's argument for the need for a service that was fast if not reliable or sequenced (to handle real-time traffic such as voice). David, can you fill in blanks here? vint (David Reed) I can fill in some blanks about splitting TCP from IP. There were a small number of proponents for specifying the Internet protocols as based on datagrams, with the idea of reliable in-order delivery being an "application layer" mechanism. This all came together at the meeting I attended in Marina del Rey. John Schoch, Danny Cohen and I each presented arguments to that effect, from different points of view. John argued based on the PUP architecture, which was an architecture based on datagrams, where streams were one option among many. Danny argued for the idea that packet speech did not want retransmission, but instead just a sequence-numbered stream of packets, where non-delivery was an option because latency was the key variable and the application could fill gaps. I argued that non-connection based computer-computer protocols, such as those we were developing for interconnected LANs, could be more general, and that end-to-end reliability was ultimately up to the endpoints to assure - for example, there were useful protocols that involved a packet from A to B, handing off of the request from B to C, and a response back from C to A were quite useful, and would be end-to-end reliable at the application level, while gaining little or no benefit from low level "reliability" guarantees. Other protocols, such as multicast, etc. were essentially datagram-oriented. I remember arguing quite strongly that you could support streams on top of datagrams, but by requiring a streams, you'd never get effective or efficient datagram services. Danny equally argued that reliable streams would create latency variability (jitter) where none was necessary. John Schoch argued that PUP was datagram-based, with streams built on top, and that architecture was quite effective. The idea of the large, randomly chosen connection identifier/sequence number was a part of the technique used in the LAN-centric internetworking protocol "DSP" (Datagram/Stream Protocol, or Dave's Simple Protocol) I had developed for MIT to allow for both datagram-like and stream-like behavior to coexist, because the connection identifier would minimize the problem of collisions in sequence space so that one need not do a 3-way handshake to authenticate an incoming message. But this was a part of a more general argument I made that related to how one could achieve idempotence in protocols - the 3-way connection setup handshake in TCP was a way to prevent delayed duplicate SYN requests from having a non-idempotent effect, i.e. doing the same command twice by accident. In many cases, I argued, idempotence is best done by the application layer, since the command itself is idempotent (i.e. the "compare and swap" instruction in a processor is inherently idempotent, because a delayed duplicate command will not compare correctly). This was one of the first end-to-end arguments, i.e. an argument that the function (idempotence) should not be done in the network layer but at a higher layer, and began my conversation with Jerry Saltzer, since he was my adviser at the time. As I recall, we 3 people, plus Steve Crocker, conspired to argue that we needed a datagram service so we could continue to do research on more general protocols, and in the heat of the argument proposed why not split out the addressing layer from the stream layer, just as we had split the TCP from the functions of the Telnet layer. (you may remember that I also was involved in changing the interaction of the TCP/Telnet layers to use a byte-oriented "urgent pointer" that was a pointer into the stream of bytes, rather than a general "process interrupt" mechanism that was being proposed, which was problematic because it embedded operating system process semantics into the network layer). In the heat of the Marina del Rey meeting, we 3 (sitting next to each other) agreed to push for splitting the TCP packet into two layers, the routing header and the end-to-end stream payload. This resulted in a sidebar meeting in the hall during the next break, where I remember it was Jon, you, Danny, me, Steve, and John Schoch, and you agreed that we should try defining how we'd split the layers, and see if the overhead would be significant. This resulted in 3 protocols, IP, TCP, and UDP (for us datagram nuts). Danny went off being happy that he could define a packet speech protocol layered on UDP, I went off happy that I didn't need to pursue DSP anymore, but could focus on how to use UDP for protocols like TFTP, which we built at MIT shortly thereafter. Vinton G. Cerf wrote: >Dave, > >Thanks a million for this rendering - I had not recalled your >involvement in the MDR event so this fills in an unintentional blank in my own recollection of this passage in Internet history. > > (David Reed) Glad to be helpful. As a young graduate student, this particular meeting was at the core of my interest, so the details are vivid in my memory. It was one of the best groups that I have ever worked with with intense argument, but a focus on constructive results. This particular interaction taught me a lot about groups, about protocol design, etc. Too bad there are so few opportunities out there for young students to participate in such processes. The "open source" community is the one area where that is still happening, and it's sad that DARPA, the current IETF, and industry don't get how to harness such things. (Reed) Pouzin was really arguing for packet networking, not free datagrams, sans connections. X.25, which only had streams, but was a packet network, was a descendant of Pouzin's work. I think he invented the term "virtual circuit", which illustrates where his head was at. He couldn't break free from the "circuit" idea, though he liberalized it a lot. This is where Gallegher and the folks at today's LIDS got it wrong, too. They defined a network as a way to multiplex a bunch of streams/circuits onto a more efficient infrastructure. They were trapped by their mindset and their model of what a network was. The real issues that bogged people down in those days were "flow control" and "reliability", whose very definitions (at least the definitions use by the networking community) were defined in terms of connections or stream or circuits. In fact, the term "congestion control" appears in no literature before the Internet design time-frame, because it was subsumed into "flow control". Though it wasn't the best idea, the idea of the "source quench" message was invented to provide a crude mechanism of congestion control that could deal with datagram-only connections, and the current TCP windowing system was initially understood as only serving an end-to-end flow control function. So inventing "source quench" was a way of breaking a controversy that surrounded the idea that gateways between networks would have no way to signal buffer overflow back to the sources. It was only later that the community (Van Jacobsen played a role, I think) invented the idea that packet drops could serve as the sensor that closed the control loop about congestion with the source. One MUST remember that we were designing the Internet protocols as*overlays* to interconnect existing networks with heterogeneous designs, but which had internal congestion control and reliability. This allowed the Internet protocols to focus on the big picture - the "end-to-end" functions. The key goal was to make the gateways simple and stateless and stupid (that was the "best efforts" insight that Vint and Bob Kahn brought to the world, which was their unique contribution. Bob often said to me that I was one of the few who really got what he meant by "best efforts"). Many people want to rewrite the history as if the Internet Protocols were designed to be implemented directly on the wire, handling in a single layer all of the problems of wireless, wired, multicast, ... transmission. In fact, the engineering approach was to accomodate MANY DIFFERENT solutions at the lower layers, and to encourage heterogeneity. Which is why I and John Schoch shared a very different viewpoint from the BBN guys, which was different from Danny Cohen. John and I were "LAN guys" who were working on networks that were inherently message-oriented, computer-computer, multicast technologies. We wanted those advantages to shine through to applications, not turned into virtual dialup telephone calls at teletype rates. Danny was a media carriage guy, who was interested in conversational packet voice over unreliable channels. To design a protocol with that diversity of undercarriage required us to resist all sorts of attempts by the "Router Guys" (like Tomlinson and McQuillen) to "optimize" for the same old ARPANET constraints (50Kb/s private lines connected by IMPs). Cisco's heritage is as Router Guys, and too many of the IETF guys these days are Router Guys. They are the new "bellheads". (Cerf) David, I think there is something incorrect about your rendition regarding Louis Pouzin. Louis was the datagram guru. The other French guy was Remi Despres and he was the one who did the Reseau Communication Par Packet - a predecessor to X.25. The latter was developed jointly by Remi, Larry Roberts/Barry Wessler/Dave Horton/and John Wedlake. When Larry Roberts was building Telenet he asked what protocols to use and I suggested TCP/IP but he rejected that claiming he could not sell datagrams and that people would only buy "virtual circuits" (sigh). Virtual Circuits were never in Louis' network world - he was all datagrams until you got to the end/end transport layer and there he introduced the voie virtuelle (virtual circuit) - not unlike TCP over IP. When another of Louis' team, Hubert Zimmerman, wrote the first OSI architecture spec, I think he had X.25 in mind as the network layer with virtual circuits built in. When I "called him" on it, he said he could not sell datagrams to the rest of the OSI community - but thought AFTER he got the X.25-based OSI specs agreed he might be able to get people to accept a connectionless addition. Eventually there was a CLNP (connectionless Network Protocol) but it was never widely implemented - nor was most of OSI except for X.400 I suppose. Remi's work had the VCs built into the network while we and Louis preferred a datagram interface and service. Oddly enough, the ARPANET had a datagram interface but an internal virtual circuit! Your comments about best efforts and the use of gateways to link networks with very different characteristics are spot on - we lost that in many respects when Cisco started routing IP directly and not encapsulating in anything other than PPP or the moral equivalent or over ethernet. I really liked the indirectness of encapsulating IP in a network layer. (Reed) Vint - you are closer to Pouzin than I ever was, so I could very well be wrong. I don't remember him ever advocating pure best-efforts datagrams without circuits at the top layer, all I remember was that he pointed out that circuit functions were better done on top of datagrams. But his normal audience was the people who used circuit abstractions (comms people), so that might have shaped the context so he would never have talked about pure datagrams as a useful capability to have. I should probably go back and read some of his stuff again, because I haven't read that literature for 25 years, and it has a way of blurring together into the OSI world, etc. Larry Roberts has always resisted the idea of moving functions out of the network, since I've known him. I remember arguing with the Telenet people repeatedly that they should think more about internetworking and less about being a homogeneous and optimized network for terminal traffic. They just never got the "Internet" idea - they were just doing the ARPANET again. They were competing with Tymnet and SNA, and seemed to think that they should move towards them. (Cerf) David, You may be correct that Louis always included VC on top of DG - he kept the VC at the edge however and not in the net. I am not sure he had applications that were purely datagram in nature -I have some old slides of his in my files from the1970s and if I ever get to it I will unearth them (I am in Hong Kong at the moment and in the middle of three weeks of travel). This is a great interchange so I'm archiving it - hope that's ok with all of you Vint (permission to post this discussion obtained from the participants) Ian Peter Senior Partner Ian Peter and Associates Pty Ltd P.O. Box 10670 Adelaide St Brisbane 4000 Australia Tel (617) 3870 1181 Fax (617) 3105 7404 Mobile (614) 1966 7772 www.ianpeter.com www.nethistory.info www.internetmark2.org (Internet Analysis Report - 2004 now available) www.theinternettapes.com (check out the new Internet history Audio CD and Ebook at this site)