[TUHS] Re: Early GUI on Linux

[Paul Ruizendaal <pnr@planet.nl>]

Thank you for highlighting that!

Several folks had already hinted at such, but your comments make clear that by 1991 the X ecosystem had come out on top in a winner-takes-all dynamic: people wanted X because that had the apps, and the apps were for X because that was the most prevalent.

This also explains that MGR on Linux was so short-lived: although it provided the terminal multiplexing that was the key use case, it did not have the application ecosystem that was apparently already important enough to motivate people to make X run on Linux very early in its existence. I had always thought of those early X applications as little more than gimmicks, but apparently they were more appreciated than I thought.


> On 27 Feb 2023, at 21:30, Dan Cross <crossd@gmail.com> wrote:
> 
> On Mon, Feb 27, 2023 at 12:22 PM Paul Ruizendaal via TUHS <tuhs@tuhs.org> wrote:
>> Thanks all for the insights.  Let me attempt a summary.
>> 
>> What it boils down to is that X arrived on Linux very early, because what the Linux hackers needed/wanted was a familiar terminal multiplexer.
> 
> While that was literally true, I think it was a little more nuanced.
> I'd perhaps put it that people wanted their familiar environments.
> Many people were used to running a lot of xterms on their
> workstations, of course, but there were other X applications people
> used regularly.
> 
> Some I remember where `gv` to look at PostScript documents, `xv` to
> view image files, `gnuplot` could generate plots in a window, `xdvi`
> to look at TeX/LaTeX output, and text editors like `emacs` had X
> front-ends that offered syntax highlighting and the like; some folks I
> knew liked `nedit`. `exmh` was my preferred email client for a while,
> and things like `xman` (was there a `tkman`, too?) was nice. Some
> system monitoring tools `xload` were considered essentials; `xbiff`
> could be useful (I could have sworn there was a `tkbiff`, too). A
> clock like `xclock` or `oclock` or something was also nice. Some folks
> liked graphical newsreaders, chat programs (I guess IRC was a thing
> back then, and I believe some `talk` client had an X front-end). There
> was a fractal viewer that I thought was fun, but I don't remember it
> much anymore (even the name...). Oh, and lots of games; I had a nice
> Solitaire version that I can no longer recall the name of. `xeyes` was
> cute, and running `xroach` was a popular (?) prank for an unsuspecting
> (but amenable) colleague.
> 
> A lot of us spent a lot of time customizing our environments, and many
> eschewed the vendor-standard environment. For instance, a lot of
> people I knew used `twm` and derivatives (`ctwm` and `tvtwm` were
> popular), and spent a lot of time tweaking menus and stuff to set
> things up the way we liked. A lot of folks also wrote custom tools
> using `tk` or `expectk`. Giving all of that up to run on Linux was a
> bitter pill to swallow, so there was a real impetus to get X running
> quickly. Personally, I kept my `tvtwm`-based environment going until I
> switched to plan9 and then to the Mac as a daily driver. I'm not sure
> I miss it, but at the time it was head-and-shoulders above anything
> you could get on Windows or (classic) MacOS.
> 
> So it wasn't just that people wanted a "familiar terminal multiplexor"
> as that people wanted the environments they had put a lot of time and
> energy into building up for themselves, and again, that often meant X.
> 
>> It seems that the pattern persists till the present day (and yes, it matches with my own dev setup/needs). I wonder to what extent this is a generational thing though. Maybe today’s twenty-somethings spend their days in front of Xcode, VStudio, Eclipse, etc. more than using multiple terminals.
> 
> I think it probably depends on what people are doing. I more or less
> switched to using VS Code for my editor, and I'm using a Mac Studio to
> write this, but my desktop is still littered with terminal windows,
> I've got a `drawterm` session open to my local Plan 9 network, and am
> logged into a bunch of old systems (Multics, TOPS-20, VMS, an IBM
> mainframe, CDC Cyber, RSTS/E, PR1ME), etc.
> 
> But the way we write software has changed pretty dramatically in the
> last 3 or so decades. I used to start with an empty C file and write
> my stuff. Things like linked-lists? Mostly implemented by hand. These
> days, there are other languages and vast collections of libraries for
> almost anything imaginable; much of what "programming" is today is
> glueing together different libraries and making them interact in
> sophisticated, often quite complex ways. I don't know that it's
> better, nor that it's always worse, but it is qualitatively different.
> So almost necessarily the toolsets and environment have changed
> accordingly.
> 
>> This ties in with another observation on early window systems. The earliest Unix window system that I could find (i.e. documented) was NUnix from 1981/82. Its desktop was designed around the idea of a dozen or so top level windows, each one being either a shell window or a graphics canvas, with no real concept of a widget set, dialogs, etc., or even of sub-windows. This paradigm seems to have been more or less the same in the Blit terminal, and carried through in MGR, Mux and even as late as 8 1/2. In the context where this serves the needs of core user group, such makes sense.
> 
> It may be instructive to look at the early X window managers in this
> regard.  One I remember was `uwm` (I think); I recall being struck
> that it reminded me of rio when I saw it.
> 
>> It is in stark contrast with developments at the lower/consumer end of the market. The original Mac, GEM and Windows all placed much more emphasis on being a graphical user interface, with standard widgets and UI design elements. On Unix and X it remained a mess. It seems that this was both for technical reasons (X not imposing a standard) and for economic reasons (the Unix wars). Linux then inherited the mess and the core user/developer demographic had no need/wish/time to fix it.
> 
> I remember the X mantra was, "mechanism, not policy." Which was fine,
> except that there wasn't much of even a default policy, which made X
> (IMHO) a bit of a bear to program and meant that interfaces were
> pretty wildly inconsistent across programs. By contrast, writing
> simple programs to draw lines on the Mac was easy.
> 
> Interestingly, frustration with this caused an almost cambrian
> explosion of new windowing environments within a few years of Linux's
> arrival on the scene. From larger efforts like Gtk (and then GNOME),
> KDE, GNUStep (which I guess might predate Linux, but not by much...),
> etc, to less ambitious things components like fvwm and Enlightenment,
> we kind of went from "OpenWindows or Motif or roll your own stuff
> around twm or something" to a whole plethora of things. It's still a
> bit of a mess, though.
> 
>> It makes me wonder when true graphical applications started to appear for X / Unix / Linux (other than stuff like terminal, clock, calculator, etc.). The graphical browser certainly is one (1993). StarOffice and Applix seem to have arrived around 1995. Anything broadly used before that?
> 
> Lots!  See above.
> 
>        - Dan C.