|
From: Charles A. <cha...@gm...> - 2015-07-23 19:06:32
|
On Thu, Jul 23, 2015 at 12:10 AM, jim s <jw...@jw...> wrote: > Subject: > Re: [multicians] Multiprocessor and/or networked Multics. > From: > jim s <jw...@jw...> <jw...@jw...> > Date: > 7/22/2015 11:45 PM > To: > do...@gm..., "Charles Anthony cha...@gm... > [multicians]" <mul...@ya...> <mul...@ya...> > Harry, > Charles, > I apologize for sending this twice. Due to "discovered" contacts for > Charles, the original message will probably only reach Harry. It has > Charles' email address corrupted with the multicians list, and the email > bounced off that thankfully, as I never post from my normal email. > > Anyway the following is an idea I had for networking and I wanted your > guy's opinion and editing before suggesting it to the main community,. > > Reply to this if you see fit. > > thanks > Jim > > I don't know if this fits with the goals of the emulation, but want to > suggest it. > > The Hercules project has been around for some time and has had to deal > with a very old system and TCP/IP for some time. The IBM systems in > question use a Channel to Channel adapter to a processor for their TCP/IP > in "real" life, so the Hercules project went with this as well as another > older method (which I don't recall at the moment). > > I am wondering if that approach could not be adapted here. The code is > not licensed such that it could be used directly, but that approach might > work here if code can be developed on the Multics system. > > The other thing that would seem to be interesting would be to see if both > emulators could support the IBM CTC adapter such that one could use a > Hercules CTCA to Multics directly. A lot of the work would be in the > underlying emulator(s) and not in the emulated Multics or Hercules code as > is currently also the case with Hercules. > > I'm adding this to the current networking thread to you guys to see if you > can see what I'm up to, and help me present it as a possibility that makes > sense to the whole of the multicians > > Having both of these work out of the box would seem to be a pretty > powerful thing that would pay dividends as far as what could be done. > > I'm not sure if the current Hercules presents in EBCDIC or ASCII natively, > but that is of course one issue off the top as far as coupling these > systems together. But that translation would take place very late in the > game, and would not be a big issue, I wouldn't think. > > the Multics implementation of course would not need any translation to > present to the emulator environment, and if we were lucky, the whole issue > would vanish anyway, as I think the IBM code is all in Ascii as soon as it > is dumped to the emulator's service process. > > Hope this makes sense enough to start a discussion. > > thanks > Jim > The underlying issue with networking as far as the emulator project is concerned is the lack of available networking code in Multics proper; both network stacks, and network aware applications. Thus, I am taking the general approach of keeping networking issues out of Multics; eg, to implement, say, a web server, I would suggest the following approach (apologies to the dps8m-developers, they've heard this): Add an FNP, configured with a full set of HSLA lines, marked as "dialup". The FNP listens on a HTTP port, and when a connection comes in, does the TCP/IP negotiation, and connects the socket to an available HSLA line, tells Multics that the line is active and sends (instead of "login Foo.Bar") "http 1.2.3.4" or the like. We patch the Multics listener code that is watching for "login", "enter", "help", etc., with a handler for "http", which starts a http server process in the same way it starts a login process. The http server process is a light-weight http server, based on some existing example, that has been ported to Multics, with all of the TCP bits thrown out. Each piece of this approach is straight-forward and manageable, it gives every appearance of a TCP/IP aware Multics without writing any Multics network stack code; and could easily be extended to other TCP/IP services, such as FTP. The biggest task is porting the server code to Multics; someone tracking down a C compiler for Multics would be a tremendous help there. -- Charles |