[ih] early networking

[John Day]

Just back to my desk,

> On Apr 30, 2024, at 17:07, Steve Crocker <steve at shinkuro.com> wrote:
> 
> John,
> 
> My understanding re the idea for the IMPs is slightly different.  Yes, there was pushback from some of the sites, but I have understood the introduction of the IMP idea came from Wes Clark.  (See https://en.wikipedia.org/wiki/Wesley_A._Clark)  He was deeply involved with small computers.  The introduction of  the IMPs into the design provided a clean separation between the subnet and the hosts.  The reliability of the overall system was dependent only on the IMPs and the lines.  In those days, it was fairly rare for an operating system to have 24 hours of continuous operation.

Yes, exactly. I was trying to keep it short. I read in Hafner’s book how Roberts was really not happy with how that meeting went and on the way to the airport Wes Clark suggested to him putting minicomputers in front of the hosts. The OSs we had were operational pretty much 24/7, except of possibly one night a week when the graveyard shift did upgrades, etc. But the idea was so that hosts would not be involved in much of he network operations.
> 
> Re the protocol, the few of us who were involved at the outset of the host level protocol design focused on connections as the basic building block.  I'll take the blame for this narrow focus if we need someone to blame.  (On the other hand, if it's a matter of giving credit, there were several of us.)  Walden's IPC concept came along after we were far down the path of designing the host-host protocol.  That said, general interprocess communication was part of our thinking from the start.

I can see how having the IMP-Host protocol create a connection from source to destination and then HHP creating the connection process-to-process made a lot of sense.
> 
> John White from UCSB also pushed for remote procedure calls as a basic building block.  It was a powerful idea but didn't dissuade us from choosing a virtual bitstream as the building block.  Maybe it should have.

Really? That early!!? Be glad you didn’t do it. IPC has to be symmetrical. We tried doing systems on top of asymmetrical models like mailboxes (Apollo) and RPC and it is always a mess. Yea, there was a big argument later over which was more fundamental RPC or IPC.  The RPC fanatics were just that. Not only was it asymmetrical, but synchronous!  When someone would bring it up, my standard response was, “O, you mean like co-routines in COBOL?” ;-) They denied but could never explain the difference.
> 
> Along a separate path, Danny Cohen was interested in real-time applications, particularly voice and flight simulation.  His work, among others, led to using messages that were not subject to flow control.

Yea, for awhile that seemed okay. But I have come to the conclusion that there should be no flows in a network that are not flow controlled. I don’t care if they are run wide open, but for security reasons there has to be a way to shut them down or at least try.
> 
> In all cases, the notion of layers was fundamental and an intrinsic part of everyone's thinking coming out of the August 1968 meeting.

Agreed, I think it is more fundamental and arises more naturally than it does in Operating Systems. In fact, the layer is a distributed resource allocator.

Take care,
John
> 
> Steve
> 
> 
> On Mon, Apr 29, 2024 at 3:58 PM John Day via Internet-history <internet-history at elists.isoc.org <mailto:internet-history at elists.isoc.org>> wrote:
>> 
>> As we all know, the ARPANET was the first major packet switching network. It was built to be a production network to lower the cost of computing for ARPA Projects.  Cyclades was built to be a network to do research on networks. Cyclades was a platform, what today would be called a clean-slate approach having seen the ARPANET. (There was considerable interaction between BBN and CYCLADES: As Dave Walden (who led IMP team) told me, so ’they wouldn’t make the same mistakes we did.’ ;-) And Jean-Louis Grangé (led the CIGALE team) spent a fair amount of time at BBN.) 
>> 
>> Because Roberts had gotten considerable resistance from the potential host sites when he proposed just a network of hosts, the IMPs were proposed to ‘off-load’ the network from the hosts. Hence for the ARPANET, the host used the IMP-Host protocol to allocate a ‘connection’ to the destination and then the Host-Host Protocol created a connection between processes in the two machines. The applications were built on top of it. (Note NCP did flow control but no retransmissions because the IMP subnet was reliable.) This is when layers were first introduced. (Again, Walden confirmed for me that there were no layer diagrams of the IMP subnet and I wouldn’t have expected them then.) This is when the idea that networking was IPC began. Walden wrote a very early RFC proposing IPC for a resource sharing network.)
>> 
>