[ih] sad news: Peter Kirstein
Craig Partridge
craig at tereschau.net
Sat Jan 11 09:32:53 PST 2020
There were X.75 gateways. CSNET used them for its IP over X.25 service
(which I think used code written by either Landweber or Comer's graduate
students).
As I recall the name of the game was to avoid an X.75 gateway -- as late as
the mid-1980s, the deployed ones all limited you to, if memory serves, a 2
packet window, whereas individual X.25 networks allowed for a bigger
window. Since CSNET was using these for international connections, the
small window caused excruciatingly slow transfers.
Craig
On Fri, Jan 10, 2020 at 11:45 PM the keyboard of geoff goodfellow via
Internet-history <internet-history at elists.isoc.org> wrote:
> seem to recall that the different x.25 networks (in different countries)
> connected via x.75 gateways?
>
> and weren't these x.25/x.75 "peering" network interconnections treated much
> like their telco mother ships voice networks in that (unlike the Internet
> where sender keeps all) there were settlements involved?
>
> furthermore, kinda recall that not only did these x.25 circuit switched
> networks clock you with a fee for the time of the virtual circuit as Jack
> mentioned, but they also summarily docked you for the bits sent as well?
>
> seem to also recall that the x.25 networks were painfully slow and didn't
> allow (more than 1?) multiple packets outstanding?
>
> and also kinda recall that the CSnet x.25 implementation employed an
> automobiles alternator type of thingy in that it opened something like 5(?)
> virtual circuits simultaneously and then sequentially round robin'd the
> data, er, packets through these vc's to effectuate higher throughputage?
>
>
> On Fri, Jan 10, 2020 at 7:53 PM Vint Cerf via Internet-history <
> internet-history at elists.isoc.org> wrote:
>
> > I had thought that there was some kind of reciprocal agreement that the
> USG
> > would charge the X.25 carrier the same amount that the X.25 carrier
> charged
> > - a kind of net zero peering arrangement. Let me do a little research.
> > v
> >
> >
> > On Fri, Jan 10, 2020 at 8:43 PM Brian E Carpenter <
> > brian.e.carpenter at gmail.com> wrote:
> >
> > > Jack,
> > >
> > > > I've wondered for
> > > > years how much of ARPA's expenses we moved into Peter's UCL budget.
> > And
> > > > whether Peter or anyone else realized what was happening.
> > >
> > > It wouldn't surprise me in the least if Peter knew and decided to keep
> > > quiet. This was presumably around 1980? The transatlantic "who pays?"
> > > discussions got very explicit when NSFnet entered the picture a few
> years
> > > later.
> > >
> > > UCL was monitoring usage as early as 1975:
> > > Monitoring and access control of the London node of ARPANET
> > > https://dl.acm.org/doi/10.1145/1499799.1499882
> > > As for many years afterwards, the bulk of traffic was from the US
> towards
> > > Europe, and most of the sessions were started by European users. (I
> > > happened to notice this paper since one of Peter's co-authors was later
> > in
> > > my team at CERN.)
> > >
> > > For proof that Peter was aware of the X.25 cost issue before too long,
> > see
> > > the bottom of page 10 at
> > http://nrg.cs.ucl.ac.uk/mjh/kirstein-arpanet.pdf
> > > "The access control was because we were
> > > incurring IPSS traffic charges on out-going traffic; the logging was
> > > because the
> > > funding agencies wanted to know how to allocate costs."
> > >
> > > Regards
> > > Brian Carpenter
> > >
> > > On 11-Jan-20 12:48, Jack Haverty via Internet-history wrote:
> > > > Thanks Vint, I hope it gives them some insight into Peter's work
> life.
> > > >
> > > > I remembered another anecdote that might also be interesting...
> > > >
> > > > Peter was the primary contact I knew on the EU side of the US/EU
> > > > partnership. I never knew the details, but I suspect there was lots
> of
> > > > diplomacy, negotiations and agreements involved involving the several
> > > > governments. Some of them of course involved money - who would pay
> > for
> > > > this new experimental Internet thing.
> > > >
> > > > Some time after the initial gateway linkage made the Internet cross
> the
> > > > Atlantic, we put a second pathway into place, by using the public
> X.25
> > > > network as just another network subsumed by the Internet.
> > > >
> > > > This was called the "VAN Gateway". After it was put into operation,
> > > > Internet traffic between the EU and US could traverse the satellite
> > > > network, that ARPA paid for, or it could traverse the international
> > X.25
> > > > network.
> > > >
> > > > The X.25 network consumed money differently than the ARPA-funded
> parts
> > > > of the Internet. In particular, the monthly cost of using the X.25
> > > > pathway was unpredictable. Since the X.25 world evolved from
> > > > traditional telephone companies, it had retained the notion of
> "calls",
> > > > and charged based on how long each call was connected. So a simple
> > > > 5-minute call would be inexpensive, while a large file transfer that
> > > > might take a day or more to complete would be quite expensive. It
> all
> > > > depended on what those pesky Users (like Peter and his crew) did as
> > they
> > > > used the Internet.
> > > >
> > > > The ARPA beancounters weren't too comfortable with unpredictable and
> > > > uncontrollable monthly expenses. So we (BBN) brainstormed about
> what
> > > > we might do to mitigate that risk.
> > > >
> > > > So....... The gateways couldn't really predict what future traffic
> > > > might appear. It depended on what those pesky Users did. The
> > > > algorithm we implemented did its best to be somewhat efficient in
> using
> > > > that expensive X.25 service. When a datagram arrived that was to be
> > > > sent over the X.25 path, the gateway would see if the "call" between
> > the
> > > > two gateways was still active from previous traffic. If not, it
> would
> > > > "dial up" the other end and then send the datagram on its way, and
> also
> > > > start, or reset, a timer. After a while when the timer expired due
> to
> > > > inactivity the gateway would simply "hang up", to prevent the
> > per-minute
> > > > charges from piling up.
> > > >
> > > > That would help, but if someone did a huge file transfer it could
> still
> > > > run up a lot of charges for ARPA. More brainstorming....
> > > >
> > > > So, ...... We simply configured the timer on the US side to be very
> > > > short - just long enough to get one datagram across the Atlantic.
> > > > probably about 1 second. We set the corresponding timer on the EU
> side
> > > > somewhat longer - a minute or so, so it would hang up quickly when
> the
> > > > user traffic subsided.
> > > >
> > > > The X.25 mechanisms behaved like the traditional telephone system,
> the
> > > > bill for each "call" would go to whichever party "dialed the phone".
> > > > With the timers set so differently, almost all of the calls of any
> > > > duration would be initiated by the gateway on the EU side of the
> > > > Atlantic, regardless of where the associated TCP connection was
> > created.
> > > >
> > > > So, ....The result was that most of the expense of using the X.25
> > > > pathway, and almost all of the unpredictability, fell onto the EU
> side
> > > > of the partnership.
> > > >
> > > > Somewhere in my education, I learned that "Management is the Art of
> > > > Moving Your Expenses Into Someone Else's Budget." I've wondered for
> > > > years how much of ARPA's expenses we moved into Peter's UCL budget.
> > And
> > > > whether Peter or anyone else realized what was happening.
> > > >
> > > > /Jack Haverty
> > > >
> > > >
> > > > On 1/9/20 8:06 PM, Vint Cerf wrote:
> > > >> thanks for this reminiscence - I am sharing with Peter's family whom
> > > >> we will see on Saturday.
> > > >>
> > > >> vint
> > > >>
> > > >>
> > > >> On Thu, Jan 9, 2020 at 5:11 PM Jack Haverty via Internet-history
> > > >> <internet-history at elists.isoc.org
> > > >> <mailto:internet-history at elists.isoc.org>> wrote:
> > > >>
> > > >> Peter was, IMHO, possibly the most important driving force
> behind
> > > >> getting the fledgling Internet to actually work! Peter (and
> his
> > > >> group)
> > > >> was the only "real" User of the Internet back in 1981. Or at
> > > >> least the
> > > >> most memorable to me.
> > > >>
> > > >> Summer/Fall 1981 was when Vint added an "Internet" task to my
> > > group's
> > > >> contract at BBN, with the assignment to make the Internet work
> > > >> reliably
> > > >> as a 24x7 communications service, just like we had been doing
> for
> > > the
> > > >> ARPANET for a decade. The Internet then was just a handful or
> so
> > of
> > > >> "gateways" (now called "routers") interconnecting networks.
> > > >>
> > > >> Most "Internet traffic" then actually moved across the ARPANET,
> > > which
> > > >> was not only reliable but also rarely did nasty things like
> > > >> dropping IP
> > > >> datagrams, reordering them, and otherwise mangling datagram
> > > >> flows. The
> > > >> neonatal TCP implementations, running over the ARPANET, didn't
> > > >> have much
> > > >> real work to do in moving users' data. The ARPANET did most of
> > the
> > > >> heavy lifting for them.
> > > >>
> > > >> However, Peter and the UCL group were actually trying to use an
> > > >> Internet
> > > >> path which involved at least 2 resource-starved gateways and
> > several
> > > >> networks of different speeds, delays, etc. Unlike most users,
> > the
> > > >> Internet was supporting Peter's group's everyday activities, not
> > > just
> > > >> occasional network experiments. And they really used it. And
> > they
> > > >> exposed lots of problems.
> > > >>
> > > >> Being on the EU side of the Atlantic, they always had a 5 hour
> or
> > so
> > > >> headstart on us every day. So there were often problems,
> waiting
> > > for
> > > >> BBN to "fix the Internet" every morning - especially after it
> > became
> > > >> known that BBN was tasked to make the Internet work as a 24x7
> > > reliable
> > > >> service.
> > > >>
> > > >> IMHO, that pressure from real users with real problems was a key
> > > >> driver
> > > >> to all the things we had to do to get the Internet out of the
> > > >> "research
> > > >> lab" to come online as a reliable communications service.
> > > >>
> > > >> Peter made (us make) the Internet work...
> > > >>
> > > >> /Jack Haverty
> > > >>
> > > >>
> > > >>
> > > >> On 1/8/20 8:09 PM, Vint Cerf via Internet-history wrote:
> > > >> > Looking at this from a different perspective, we had Peter in
> > > >> our midst and
> > > >> > contributing heavily to networking successes and spread from
> > > >> about 1967
> > > >> > when I first met him while at UCLA. That's over half a
> century.
> > > >> What a
> > > >> > gift! His work is still evident and his story of collaboration
> > > >> undiminished
> > > >> > by his departure from our midst. Yes, another giant has left
> us,
> > > >> but his
> > > >> > work remains to remind us of what we can do when we work
> > > >> together. None of
> > > >> > that is gone though we shall not see him again in this world.
> > > >> >
> > > >> > I am glad to have called him "friend" for many, many years.
> > > >> > vint
> > > >> >
> > > >> >
> > > >> > On Wed, Jan 8, 2020 at 9:27 PM Bob Hinden via
> Internet-history <
> > > >> > internet-history at elists.isoc.org
> > > >> <mailto:internet-history at elists.isoc.org>> wrote:
> > > >> >
> > > >> >>
> > > >> >>> On Jan 8, 2020, at 11:38 AM, Toerless Eckert via
> > > >> Internet-history <
> > > >> >> internet-history at elists.isoc.org
> > > >> <mailto:internet-history at elists.isoc.org>> wrote:
> > > >> >>> On Thu, Jan 09, 2020 at 08:26:00AM +1300, Brian E Carpenter
> > via
> > > >> >> Internet-history wrote:
> > > >> >>>> This is really bad news. And don't forget his team's true
> > > >> pioneering in
> > > >> >> video-conferencing over the Internet too.
> > > >> >>> Indeed
> > > >> >> Yes, very sad news indeed. Sigh...
> > > >> >>
> > > >> >> Bob
> > > >> >>
> > > >> >>> All MICE and friends will miss him dearly, and keep fond
> > > memories.
> > > >> >>>
> > > >> >>> Toerless
> > > >> >>>
> > > >> >>>> Regards
> > > >> >>>> Brian Carpenter
> > > >> >>>>
> > > >> >>>> On 09-Jan-20 06:17, Vint Cerf via Internet-history wrote:
> > > >> >>>>> I am sorry to relay the sad news that Peter Kirstein
> passed
> > > >> away this
> > > >> >>>>> morning (London time). He was a key implementer and
> promoter
> > > of
> > > >> >> networking,
> > > >> >>>>> participating in both the ARPANET and Internet
> developments
> > > >> as well as
> > > >> >> the
> > > >> >>>>> UK Coloured Book and Open Systems Interconnection
> protocols.
> > > >> >>>>> https://en.wikipedia.org/wiki/Peter_T._Kirstein
> > > >> >>>>>
> > > >> >>>>> vint cerf
> >
>
> --
> Geoff.Goodfellow at iconia.com
> living as The Truth is True
> http://geoff.livejournal.com
> --
> Internet-history mailing list
> Internet-history at elists.isoc.org
> https://elists.isoc.org/mailman/listinfo/internet-history
>
--
*****
Craig Partridge's email account for professional society activities and
mailing lists.
More information about the Internet-history
mailing list