Talk:Time-sharing system evolution

Is MS Windows a timesharing system?

I've always thought a timesharing system had to support multiple concurrent users. AFAIK no Windows version does this - the NT-2000-XP-Vista lineage allows multiple logged-on users, but only 1 can use the system interactively at a time (I'm not sure about background tasks under multiple logged-on users). Philcha (talk) 20:41, 16 December 2007 (UTC)

Actually some Windows versions have supported this as far back as 1998, with Windows NT 4.0 Server, Terminal Server Edition. The current implementation is called Remote Desktop Services. But, MSDOS never provided any such thing. Snori (talk) 10:59, 1 June 2014 (UTC)

MS Windows other than 3

I think the table (which is an excellent summarising device) understates the complexity of the ancestry of Windows versions beyond Windows 3 (which was just a GUI front-end for MS-DOS):

• Win NT was mainly influenced by OS2 and Vax VMS, and not significantly by Win 3 except that they support MS-DOS commands. Win 2000, Win XP and Win Vista are descendants of NT, with increasing imports of "user-friendliness" from the Win 95 lineage (earlier) and Mac OSs (recently).
• I'm less sure about the main influences on Win 95, 98 and ME. They use much the same UI design and the same shortcut format (in the Start Menu, etc.) as the NT-2000-XP-Vista lineage, and like NT etc. support long filenames; but they resemble Win 3 in using a FAT-based file system and relying on co-operative multi-tasking because they lack pre-emptive multi-tasking.

The problem is how to express all this:

• A full network of relationships / influences could be as confusing as a spider's web. It might help to colour-code the lines from each source / predecessor.
• One could distinguish major and minor influences by the width of connecting lines in a network, but that would rather subjective. Philcha (talk) 20:37, 16 December 2007 (UTC)

As far as I know, Windows 9x/Me ("Windows OT") does preemptive multitasking of 32-bit programs.

But is this about time-sharing systems, or is it about systems that provide interactive computing that allows a user to have a session with a computer, whether the computer in question supports multiple users or not? See, for example, the mention of LISP and Smalltalk machines. If it's about the former, then, whilst NT had underlying support for time-sharing, eventually use for Windows Terminal Server, "Windows OT" never had that. Guy Harris (talk) 21:23, 16 August 2024 (UTC)

Nonsense that has to go

I struggle to understand how any of these ever got into this article.

Programming languages:

• APL
• LISP
• Smalltalk

Yes, all these ran on time-sharing systems, but so did BASIC, umpteen editors and payroll programs etc.

Single task "OS"s:

• CP/M
• MSDOS

These couldn't run two apps at a time, let alone for remote users.

Possible entries:

• OS/2 (with the addition of Citrix MULTIUSER and Winview)
• Microsoft Windows (but only from Windows NT 4.0 Terminal Server Edition)

I will remove them shortly unless anyone can provide a compelling argument against doing so. Snori (talk) 11:22, 1 June 2014 (UTC)

Done. I have left Windows and OS/2 for now, but will re-word in line with my comment above some time in the next few days. Snori (talk) 05:26, 4 June 2014 (UTC)

That just shows your ignorance. Read corresponding articles for that languages. They had early implementations which didn't depend on other OSes. In fact APL\360 was almost the first successful time-sharing option in time where IBM didn't offer any comparable alternative. And that implementation had very weak dependency on its host OS. Lisp machines even had their own hardware. These early systems influenced time-sharing development up to this day. 188.208.99.37 (talk) —Preceding undated comment added 11:16, 18 January 2019 (UTC)

Yes, I have the impression that APL\360 supported multiple users on multiple terminals on a single machine by doing its own time slicing, rather than by running multiple instances of the APL interpreter as ordinary programs atop a time-sharing OS. (Later versions may have left the time slicing/multi-user work up to the OS.)

What LISP systems did their own time-sharing in that fashion? As the edit comment for the edit that removed LISP from the list said, "Lisp machines were single user workstations, not time-sharing", and those systems weren't "early" in the history of time-sharing.

The same applies to Smalltalk. Guy Harris (talk) 17:20, 18 January 2019 (UTC)

Unfortunately, I'm not an expert on Lisp and Smalltalk. But I didn't rush to remove them either. OTOH, APL\360 source code is available at Computer History Museum and was fixed to run under Hercules emulator and OS/360, so it's known for sure that in order to do time-sharing it made invasive changes to its host OS which didn't provide any time-sharing on its own. I've restored APL with a cite to a paper which describes that period of history.188.208.99.37 (talk) 17:36, 18 January 2019 (UTC)

If this article is about time-sharing, it should only mention multiuser systems, and should only mention languages in the context of a system that provides time-sharing with support for that language. Thus, just as it mentions DTSS rather than BASIC, it shouldn't mention APL, but could mention the APL\360 system, as the time-sharing was provided by APL\360's "time-sharing supervisor". This paper from IBM suggests that the original (internal-to-IBM) APL\360 was "a dedicated system"; this paper indicates that the version it described ran atop a modified version of DOS/360, including its own "time-sharing supervisor". It should not include LISP or Smalltalk workstations, however, as they didn't do time-sharing.

If it's about interactive computing in general, it can include various single-user interactive systems - which, I think, date back to the 1960s as well. Guy Harris (talk) 22:01, 16 August 2024 (UTC)

DTSS

DTSS was certainly not a DEC system as it ran on GE (later Honeywell) hardware.

I was not completely comfortable adding it to the MULTICS/UNIX section, but that seemed better than the alternatives. The DTSS and MULTICS projects overlapped. Doug McIlroy told me in an email thread, "Though the Multics team was well aware of DTSS, I can't think of any specific influence from DTSS." LCarl (talk) 14:38, 12 May 2019 (UTC)

One alternative is "give DTSS its own section"; is putting it in the Multics section (which is now, appropriately, separate from the UNIX section - Multics was a strong influence on some aspects of Unix, but there are also significant differences) better than giving it a section of its own? Guy Harris (talk) 22:37, 16 August 2024 (UTC)

LTSS/CTSS

I do not have access to reliable sources, but the Livermore Time Sharing Service (or System) probably should be included on this list. I believe it was written in a variant of Fortran. LTSS was a forerunner of the Cray Time Sharing System, an operating system used on some early Cray-1 systems. CTSS had a unique file deletion command, DESTROY. 207.231.0.246 (talk) 14:34, 5 July 2024 (UTC)

LTSS has a page - Livermore Time Sharing System - but it has few references. (And be careful about calling the Cray Time Sharing System "CTSS", as it might cause confusion with the original CTSS.) Guy Harris (talk) 22:32, 16 August 2024 (UTC)

I'm not sure Linux should have a table entry of its own

The "UNIX and derivative systems" entry covers it, as Linux (at least in its "GNU/Linux" form, with GNU and other userland components) is a derivative in the programming and user interface sense, even if it's not a derivative made directly from UNIX source code (at this point, even the commercial UNIXes have probably added or changed as much code, especially in kernel-mode code, as they inherited from Bell Labs/AT&T UNIX, and that's probably even more the case with the *BSDs). I'd be inclined to have that entry mention Linux - e.g., "derivative systems, including Linux".

(In addition, most UNIX derivatives, including Linux, are running either on embedded devices/appliances, personal computing devices/workstations, or servers; to the extent that they're used for time-sharing, that's probably mostly people SSHing in to perform various tasks. Linux, and other current UNIX derivatives, are "time-sharing systems" mostly by history and by not having lost their time-sharing capabilities, not by being used as an OS for a classic time-sharing system.) — Preceding unsigned comment added by Guy Harris (talk • contribs) 23:01, 16 August 2024 (UTC)

GE/Honeywell Timesharing is missing

GE (later bought by Honeywell) had a notable role in timesharing in the 60s-70s that isn't mentioned here. The first DTSS was built by Dartmouth on the GE265, which was a bundle of a GE235 "mainframe" and a GE DATANET-30 front-end terminal multiplexer. GE itself commercialized this system for a time.

In parallel, while Dartmouth migrated DTSS to the GE635, GE itself built a separate timesharing subsystem for the GE635 running the GECOS batch OS. This was marketed (first by GE, later by Honeywell) at least into the 1980s. I don't know how many installations there were, but it was substantial for the time. When Bell Labs at Murray Hill dropped out of the Multics coalition, they converted their GE645 Multics computer back to a GE635, and ran the GE timesharing system on it. (This was concurrent with the early development of Unix on the PDP-6 there.)

GE also marketed Multics into the 80s. I don't know how many sites there were, perhaps around ten.

I think this deserves a mention. PHK46 (talk) 16:25, 18 July 2026 (UTC)

"Missing" in what sense?

The article has

• a lead, which doesn't cover any particular systems or organizations' contributions to the evolution of time sharing;
• Time-sharing system evolution § Time-sharing, which also doesn't cover that - it mainly talks about communication with time-sharing systems;
• Time-sharing system evolution § Family tree of major systems, which has a "GE/Honeywell mainframes" item that covers DTSS and Multics.

The one thing missing is the time-sharing subsystem for GECOS. That would presumably go into the "GE/Honeywell mainframes" section. Guy Harris (talk) 22:12, 18 July 2026 (UTC)

And add it to the table in Time-sharing system evolution § System descriptions and relationships, which also already mentions DTSS and Multics. Guy Harris (talk) 22:19, 18 July 2026 (UTC)

(I'm totally new to commenting here, so sorry if I do it improperly.)

Yes, the GECOS Timesharing subsystem deserves to be mentioned in the table entry for GE/Honeywell mainframes. It can be considered a descendant of DTSS. My understanding is that it was created concurrently with the migration of DTSS from the GE265 to GE635. (But I don't know actual dates.) But it wasn't very influential on future developments.

I am dubious that the GE 265, initial platform for DTSS, can be considered a mainframe. But that may be insignificant to the point of the article. PHK46 (talk) 16:42, 19 July 2026 (UTC)

I am dubious that the GE 265, initial platform for DTSS, can be considered a mainframe. So perhaps it should be "GE/Honeywell hardware" or something such as that? Guy Harris (talk) 18:53, 19 July 2026 (UTC)

That seems fine. PHK46 (talk) 19:25, 19 July 2026 (UTC)

Another thought - it seems like the PDP-1 timesharing system might deserve a mention here. Checking the timelines, it seems that it was just slightly behind CTSS in development. It was sort of a "toy" system, and I don't think it had much influence on subsequent systems. But it was certainly influential due to Space War.

(If you haven't figured it out yet, I have first hand knowledge of these things.) PHK46 (talk) 19:46, 19 July 2026 (UTC)

So perhaps it should be "GE/Honeywell hardware" or something such as that? I went with "GE/Honeywell systems", paralleling "DEC systems". Guy Harris (talk) 20:06, 19 July 2026 (UTC)