How X Window Managers Work, and How to Write One (2014)

I made an X10 window manager based on "uwm" in 1987, which had pie menus, and was extensible and scriptable in FORTH!

https://donhopkins.com/home/archive/piemenu/uwm/

https://donhopkins.com/home/archive/piemenu/uwm/Menu.c

https://donhopkins.com/home/archive/piemenu/uwm/fuwm-main.f

I didn't see a bright future for X (and still don't ;) ), so I switched to NeWS, and began writing pie menus and window managers in PostScript.

NeWS (which was James Gosling's creation before Java and after Emacs) was architecturally similar to what is now called AJAX, except that NeWS coherently:

- used PostScript code instead of JavaScript for programming.

- used PostScript graphics instead of DHTML and CSS for rendering.

- used PostScript data instead of XML and JSON for data representation.

Pie Menus and Tab Windows for NeWS in object oriented PostScript:

https://donhopkins.com/home/archive/piemenu/pieui/

NeWS Tab Window Demo:

https://www.youtube.com/watch?v=tMcmQk-q0k4

At Sun, we even used the above code to write an X11 ICCCM "Open Window Manager" in PostScript for X11/NeWS, which wrapped your X11 windows in NeWS tabbed frames with pie menus, implemented in PostScript, the same as all your NeWS windows used, all running in the same address space, and easily centrally customizable and extensible.

https://news.ycombinator.com/item?id=13817649

>At Sun we experimented with implementing an X11 window manager in NeWS. We didn't have transparency at the time (1992), but we did support shaped windows!

>The NeWS window manager supported cool stuff (for both X11 and NeWS windows!) like rooms, virtual scrolling desktops, tabbed windows, pie menus, was easily extensible and deeply customisable in PostScript, and ran locally in the window server so it could respond instantly to input events, lock the input queue and provide feedback and manipulate windows immediately without causing any context switches or dealing with asynchronous locking, unlocking and event handling. You'd never lose a keystroke or click when switching between applications, for example.

https://news.ycombinator.com/item?id=22456153

>And we had a lot of internal discussion about how NeWS fit into Sun's window system strategy. We developed a prototype X11 window manager in NeWS, to prove how much better NeWS can handle seamlessly integrated NeWS and X window management much better than X can manage its own windows. The next step we wanted to take was to write a user-extensible HyperCard-like window manager using HyperNeWS/HyperLook. But Sun management wasn't having it. They actually wanted to do the worst-possible upside-down solution and put NeWS applications inside of X-Windows managed by OLWM, precluding the possibility of arbitrarily shaped windows, tabbed windows, pie menus, all stuff we'd been doing for years with NeWS that we'd have to give up in the name of X interoperability, after we'd already proven we had a working better solution with "owm".

https://donhopkins.com/home/archive/NeWS/owm.ps.txt

>% Open Window Manager

    /ClassX11ManagerMixin ClassWindow
    [/IconWindow /WidthInc /HeightInc /FirstMapping]
    classbegin

        /NewInit { % client => -
            /NewInit super send
            self /WM_STATE xdeleteproperty % The frame is not a ICCCM window
        } def

https://www.donhopkins.com/home/archive/NeWS/sevans.txt

>Don> If you don't want to dump NeWS, then why have you been caused us so much trouble?

>Steve> I could ask you the same question, "If you want NeWS to be a commercial success, why has NeWSTech been so subborn in the sense of resisting trying to fit into the X environment."

>Don> That's not the same question. We want NeWS to be a commercial success, but "commercial success" is not a standard defined by the X Consortium. I think OWM can do a beautiful job of fitting the X environment into NeWS. If you find that concept terrifying, then you know how we feel about the inverse, knowing that we will have to give up many goals we designed for and successfully achieved, in order to accomodate a half assed "fallback" solution to satisfy some of our customers who want to run our competitors' software (because we aren't allowed to make our software good enough for them to want to run).

>We have had a great deal of trouble trying to fit into the X environment. It has taken a huge amount of PostScript code to deal with many X interoperability issues ranging from undocumented selection protocols that XView and OLIT can't even agree on, to input focus grabbing kludges to work around OLWM's incorrect focus tracking behavior, to the fullscreen.ps hack to keep X from grabbing the pointer when NeWS is tracking it. Then there are problems we could do nothing about, like the system locking up when OLWM grabs the server (grabbing the pointer after grabbing the server, and not checking the return value). Many of these problems should have been addressed by fixing X programs or the server, but they were not, instead we had to code around them in PostScript when we could.

NeWS had its own problems, of course, but its biggest problems weren't technical, but political, and it wasn't free despite all the hard effort, spilled blood, and broken promises. But the consequences of that experience did help to make Java free, eventually. (But then Oracle later ruined worse than Sun ever could have.)

https://news.ycombinator.com/item?id=22457490

>James Gosling fought very hard for NeWS, but in the end failed to convince Sun to do the right thing. He was optimistic when he talked me into going to Sun to work on NeWS after I'd already given up on it by 1990, saying that Sun had turned over a new leaf, and was soon going to announce their commitment to NeWS: [...]

>"We'll get it out, even if I have to spill some real blood on the floor." -James Gosling, on freeing NeWS, March 6, 1990

>[...] It has been very clear for a long time that DEC explicitly targeted NeWS as something to be trashed. That was probably the major reason for giving NeWS such a ridiculously low profile. There were folks at sun who didn't want to give DEC a bigger target. The folks running the show now have more guts.

>James.

https://www.donhopkins.com/home/archive/NeWS/flame.txt

>November 11, 1990:

>I have been hoping for Sun to make Open Windows free since it was called SunDew, and during that time, I've made it perform many indescribable acts (both on and off stage), worked with quite a few companies trying to make it a succeess, and drained much much more of my energy than I ever knew I had to give into that incredible piece of software.

>I have been following the messages on the network in the aftermath of the OWPS "free for $1000" disaster... The big problem was not the $1000. It was the word "free".

When your solution to "make something secure" is to isolate instances of it in airtight sandboxes, IT IS NOT SECURE.

Theoretically Xorg can be made fully secure: just isolate clients so they can only receive events and bitmap information from windows created on the same client connection. It would be relatively straightforward, if quite involved, to implement.

But nobody wants to implement it because everyone qualified to do so has jumped ship to Wayland. The X architecture is so fatally flawed that the most straightforward way to fix it is to start from scratch, and that's what Wayland is.

X is like global warming: one hundred percent of the people who are in the least wise knowledgeable agree that it is a problem. Unlike global warming, however, that problem has a fix: Wayland.

So just... shut up with the irrelevant bullshit and use Wayland, like all the Linux graphics maintainers and distro maintainers want you to do and have been telling you to do for years now... or find your shit unsupported.

Your toolkit should abstract all that away. We have these wonderful things called dynamic libraries, that X was designed for a world without because they didn't exist under Unix yet.

The last time I used wlroots, via sway, I encountered a bug where I experienced 'multicolored snow' when watching a video in firefox. I wanted to record a video of it to report a bug, but of course that wasn't possible because screen sharing / recording still isn't well supported. Back to gnome I went, and I still don't think wayland is ready for primetime on smaller window managers

The effort to improve is welcome and appreciated.

Casting multi-use graphical computing aside is unwelcome.

Most of the toxicity centers on HOW that discussion has played out.

To be fair, many users coming up on single user graphical computing have no idea what the problem is, do not have use cases and we all know the rest.

The users who do understand all that are pissed. They are being told none of it really matters or is necessary, and so on...

Of course all that is definitely not appreciated however spiffy watching videos about it all may be.

And, those users totally get the need to see the embedded use case improvements get done, and still are being asked to just forget multi user graphical computing was ever a thing because reasons...

So yeah, here we are.

People will care about Weyland exactly as much as the Weyland team cares about them.

Multi user graphical computing aware users are not cared about at all despite their repeated explanations on how they depend on that capability. Depend, as in, not having it becomes very expensive for them. Expensive enough to wash away all the value they are being told matters more to everyone else.

Non multi user graphical computing aware people basically just want it settled so they can have few overall worries. They feel some care aimed their way.

The embedded people are super happy and are cared about a lot.

All adds up to a very toxic state of affairs.

Factoring it all down:

Did we absolutely have to trade multi user graphical computing away?

Should the answer be no, this whole ugly mess will go away.

Continuing with a yes means a very painful and drawn out mess for years to come.

And it will be that way because there are plenty of people who really do get a lot of value out of multi user graphical computing.

Who knew?

I bet the former X devs did. And they just do not care.

Why exactly should they get any back in return?