[ih] Origination date for the Internet

[Guy Almes]

Vint,
   That's the odd thing: it *was* a single (Cisco 7xxx) router.  Hence 
the obvious problem I noted.
   With Paul Vixie configuring the router, I suspect that what could be 
done was done.  But I am note aware of any early-90s router technology 
that would fix the problem I noted.
   I'd be interested to hear from Paul or anyone else on what the CIX 
founders planned to do about this problem.  Several of them were 
technically very savvy and would have anticipated the situation/problem.

   This contrasts, of course, with the 1995-era Chicago NAP -- 
implemented with an Ameritech ATM switch.  In that case, using ATM PVCs, 
they were able to do layer-2 switching with both local and long-distance 
circuits.

   By the way, your note shows that you are now having a reaction 
similar to the one I had back in '91.
  	-- Guy

On 10/29/10 2:20 AM, Vint Cerf wrote:
> Guy,
>
> You are making an interesting assumption I think. The way I understood
> CIX is that it was an ethernet with three routers, one each operated
> by PSI, UUnet and CERFnet. There isn't a unique "CIX router". There
> are three of them. Does that help?
>
> v
>
>
> On Thu, Oct 28, 2010 at 10:21 PM, Guy Almes<galmes at tamu.edu>  wrote:
>> John,
>>   My question was really about the original PSI/ UUnet/ CERFnet design and, I
>> guess to make it interesting, you'd have to posit a 4th member, just say
>> FOOnet.
>>   Suppose PSI and UUnet each have routes to a given destination D that does
>> not involve going through the CIX router.
>>   Suppose CERFnet and FOOnet connect to the CIX router and need to get to D.
>>   Suppose, further, that CERFnet would prefer to get to D via PSI and that
>> FOOnet would prefer to get to D via UUnet.
>>   What would be the routing table entry in the CIX router for D?
>>   How would "bilateral" agreements help?
>>   This has puzzled me for almost 20 years,
>>         -- Guy
>>
>> On 10/28/10 8:45 PM, John Curran wrote:
>>>
>>> Guy -
>>>
>>>    Routing policies between peers over the CIX were agreed to on
>>>    a bilateral basis, just as most peering is done today. The only
>>>    exception I know of was due to the CIX/ANS interconnection and
>>>    combits quandary, which resulted in ANS being a transit network
>>>    for select NSF regional network who joined the CIX.  Paul Vixie
>>>    was the network engineer configuring the CIX router over most of
>>>    its operational life, and could supply the specific details...
>>>
>>> /John
>>>
>>> On Oct 28, 2010, at 8:59 PM, Guy Almes wrote:
>>>
>>>> Richard,
>>>>   Right.
>>>>   The original CIX was in 1991 and, interestingly, done as a router rather
>>>> than a switch.  And with a T1 circuit coming from each participant.  I
>>>> always wondered how PSI, UUnet, and CERFnet agreed on routing policies.
>>>>    -- Guy
>>>>
>>>> On 10/28/10 7:37 PM, Richard Bennett wrote:
>>>>>
>>>>> 3Com was founded way back in '79, but it took Ron Crane a while to
>>>>> figure out the black brick; I think their Multibus adapter was about
>>>>> '81. The first single-chip Ethernet controller, the SEEQ 8001, didn't go
>>>>> into production until '83.
>>>>>
>>>>> BTW, as far as I can tell, the first CIX was a Cisco 7500 that connected
>>>>> PSI, UUNET, and Cerfnet somewhere around McLean, VA in 1991. It was
>>>>> moved to Palo Alto shortly afterward.
>>>>>
>>>>> PAIX came long in 1996 as a carrier-neutral NAP alternative.
>>>>>
>>>>> RB
>>>>>
>>>>> On 10/28/2010 3:25 PM, Vint Cerf wrote:
>>>>>>
>>>>>> beats me - 3COM was in operation by then and Berkeley BSD 4.x had also
>>>>>> been released, right?
>>>>>>
>>>>>> v
>>>>>>
>>>>>>
>>>>>> On Thu, Oct 28, 2010 at 6:19 PM, Guy Almes<galmes at tamu.edu>    wrote:
>>>>>>>
>>>>>>> Vint et al.,
>>>>>>> I wonder about how many (mostly departmental) LANs were running
>>>>>>> TCP/IP and
>>>>>>> connected to the ARPAnet by 1-Jan-83?
>>>>>>>
>>>>>>> -- Guy
>>>>>>>
>>>>>>> On 10/28/10 4:44 PM, Vint Cerf wrote:
>>>>>>>>
>>>>>>>> actually ISI tracked TCP/IP capability during 1982; the primary
>>>>>>>> regular use was from Europe, especially the UK, prior to january
>>>>>>>> 1983;
>>>>>>>> by then there LANS connecting to the ARPANET by way of gateways
>>>>>>>> (Proteon was around with its rings - Noel Chiappa is that correct?).
>>>>>>>> Then came Cisco but i guess after 1984.
>>>>>>>>
>>>>>>>> Of course during 1982 many ARPANET sites came up on TCP/IP in
>>>>>>>> parallel
>>>>>>>> with NCP.
>>>>>>>>
>>>>>>>> v
>>>>>>>>
>>>>>>>>
>>>>>>>> On Thu, Oct 28, 2010 at 4:08 PM, Miles Fidelman
>>>>>>>> <mfidelman at meetinghouse.net>    wrote:
>>>>>>>>>
>>>>>>>>> Bob,
>>>>>>>>>
>>>>>>>>> Bob Hinden wrote:
>>>>>>>>>>
>>>>>>>>>> I still have my "I Survived the TCP Transition 1/1/83" red button.
>>>>>>>>>>
>>>>>>>>>> In my view this was the time when the Internet became operational
>>>>>>>>>> as
>>>>>>>>>> people starting using it for their day to day work, instead of a
>>>>>>>>>> set of
>>>>>>>>>> researchers. Conception and birth occurred earlier :-)
>>>>>>>>>>
>>>>>>>>> Actually, that raises another interesting question: At what point,
>>>>>>>>> prior
>>>>>>>>> to
>>>>>>>>> 1/1/83, if any, was there a minimal set of networks, gateways, and
>>>>>>>>> end
>>>>>>>>> systems that were passing IP packets on an ongoing basis - as
>>>>>>>>> opposed to
>>>>>>>>> being cobbled together to run some experiment or other, and then
>>>>>>>>> brought
>>>>>>>>> back down? Can we isolate a date when IP packets started flowing and
>>>>>>>>> never
>>>>>>>>> stopped?
>>>>>>>>>
>>>>>>>>> Miles
>>>>>>>>>
>>>>>>>>>
>>>>>>>>>
>>>>>>>>> --
>>>>>>>>> In theory, there is no difference between theory and practice.
>>>>>>>>> In<fnord>    practice, there is. .... Yogi Berra
>>>>>>>>>
>>>>>>>>>
>>>>>>>>>
>>>>>>>>
>>>>>
>>>
>>>
>>
>