Talk:Interface Message Processor

IMP varieties

More IMP varieties? 316, Pluribus, C30. Larsbrinkhoff (talk) 06:38, 16 March 2018 (CET)

As in, I should say more about them? :-) Jnc (talk) 14:45, 16 March 2018 (CET)

Not necessarily. It's just that current text makes it sound like the 516 IMP was the only model. Larsbrinkhoff (talk) 14:52, 16 March 2018 (CET)

Huh? "Later, other similar minicomputers were also used." Jnc (talk) 15:24, 16 March 2018 (CET)

Sorry, I missed that! Larsbrinkhoff (talk) 15:57 16 March 2018 (CET)

Eh, no biggie. It's not that clear, so I'll add a bit of detail. Jnc (talk) 16:06, 16 March 2018 (CET)

RT IMPs

I was under the impression that IBM RTs ( weird old RISC ) served as a IMP*, but apparently they predated IMPs,

*IMP is a misnomer, They were passing computer to computer message packets, which would later become part of IP, Internet Proticall

"The National Science Foundation Network (NSFNET) was the forerunner of the Internet. From July 1988 to November 1992, the NSFNET's T1 backbone network used routers built from multiple RT PCs (typically nine) interconnect by a Token Ring LAN."

ForOldHack (talk) 09:53, 5 July 2020 (CEST)

A few points on what ForOldHack wrote here:

• IMPs first appeared in 1969 and the IBM RT came out in 1986 so IMPs predated RTs by some 17 years. While it's possible that there were some later IMPs that used RTs, I'm highly doubtful. I only used computers connected to IMPs and ported an IMP driver from the PDP-11 to the microVAX so I didn't use all possible IMPs, but by the time the IBM RT came along that style of building computer networks had long passed. We were now using local-area networks and point-to-point links with routers (and were just making the change from calling them gateways to calling them routers).

No, RT's were never IMPs, although they were routers in the IBM/UMich NSFNet. Jnc (talk) 03:03, 26 November 2020 (CET)

• I don't know about IMP being a misnomer but that is what they were called. Yes, they passed computer message packets (usually just called packets) and some of those packets were Internet Protocol packets but saying that the packets they passed later became part of IP is a very strange way to look at it. That's just not how it worked. The IP protocol was defined, it was implemented in IMPs (and other places), and then IMPs forwarded IP packets as well as all the other stuff they did (NCP and routing packets and I don't know what all else).

Actually, the packets passed from host to host via IMPs were called 'message's; those could be broken up into multiple smaller 'packet's by the IMP, but were re-assembled into the original message before being delivered to the destination host. 'Message's were packets, in the generic meaning of that term, but in the ARPANET world, 'packet' referred exclusively to the fragments sent between IMPs.

IMPs didn't know anything about IP (it's not clear if that "IMPs forwarded IP packets" meant that, I cheerfully admit). IP packets sent over the ARPANET were wrapped in Host-to-IMP Protocol, which the IMPs understood, but the IP packets inside were just a pile of bits to the IMP. TACs (the later evolution of the original TIPs of the ARPANET) did understand IP, but there were pure hosts, unlike TIPs, which also functioned as IMPs. (Physical TIPs were converted to TACs, if this whole deal isn't already confusing enough; see here for details.) Jnc (talk) 03:03, 26 November 2020 (CET)

• I started working on the Internet in 1982, several years before the NSFNet came along in 1988, and I was hardly there at the beginning. In fact, I figure I was about the third generation working on the Internet. In other words, the NSFNet was not in any way the forerunner of the Internet. The NSFNet was a transition away from the ARPANet being the backbone of the Internet and it soon gave way to the commercial Internet backbone that grew up in the 90s.

Dab (talk) 21:48, 25 July 2020 (CEST)