Thanks, Gwern. So your line of interest is similar to mine. :) I have one question: the page you referred to says "live hacking [in stumpwm] means you can hose your X session" It's correct, and I have done it several times in Sawfish, like infinite loops. (I've never tried StumpWM.) But isn't it the same for any of these WMs that if you have errors in your code, your WM can be almost unusable, no? Or is it less likely in haskell? It's better to byte-compile in Lisp, too. Sawfish's lisp (librep) is poor, but I guess common lisp compiler may be able to emit good warnings. (And please make it habit to record the date of last web page edit, which is really informative for readers.) I don't understand why large desktop is not so much common (famous example which has LD is fvwm); it's really convenient since geometric meaning is intuitive. For example, my top-right viewport (a viewport is the portion shown in a monitor) is assigned for browser, and I switch the viewport by Shift + cursor. I only have one workspace, but some Sawfish user have several large workspaces. # Wow, you're interested in 柄井川柳 (Karai Senryu)? I don't know # about him at all, but I'm interested in Edo period senryu, too. # I'm Japanese. :) Teika (Teika kazura, retired Sawfish developer)

On Tue, 6 Sep 2011 04:02:39 -0400, Brandon Allbery wrote:

...

I don't understand why large desktop is not so much common[...]

Window managers don't do it any more because X11 has had the ability to do it itself for some 10 years now (see the "Screen" section in xorg.conf, notably the "Viewport" declaration).

If I understand correctly, they're different. X server resolutions are fixed by the video card, and you can only choose one from them. E.g. in my case 1280x800, 800x600, etc. And X "viewport" happens if the _server_ resolution is bigger than the _monitor_ resolution. Large desktop is provided by window managers, and there's no limitation. It can be 10000x10000 or bigger. Unfortunately Sawfish's simple implementation limits it to 32767, which is a realisitic problem if you have several monitors. Thanks anyway. Teika (Teika kazura)

On Tue, 13 Sep 2011 11:07:04 -0400, Brandon Allbery wrote:

...

On Tue, 6 Sep 2011 04:02:39 -0400, Brandon Allbery wrote:

...

...

I don't understand why large desktop is not so much common[...] [...] While it was originally introduced to allow multiple monitor resolutions to coexist

Probably you're talking about RandR, and you're right. Another unfortune is that it's (also) poorly documented. Though I'm not sure if meddling RandR to achieve large desktop is the best way...

(There is an additional limitation based on protocol pixel addressability ― +/- 32767, IIRC ― but I think modern servers have extensions to correct that

16-bitness is part of X11 protocol (Xlib functions do take 32 bits arguments, but don't get fooled) and it persists. "X12" page (http://www.x.org/wiki/Development/X12) says: 15 bit coordinate limit Ouch. 32768 pixels at 100dpi is 8.3 meters. Regards, Teika (Teika kazura)

Thanks a lot for your detailed explanation. If it's that problematic, it's peculiar that EWMH (Extended window manager hints) doesn't deprecate large desktop, whose latest being 1.3 or 1.4-draft2, updated 2005 or 2008. Anyway ewmh is deserted, and gnome and kde people don't care others any more. ;) Teika (Teika kazura)

2011/9/6 Teika Kazura :

I have one question: the page you referred to says "live hacking [in stumpwm] means you can hose your X session"

It's correct, and I have done it several times in Sawfish, like infinite loops. (I've never tried StumpWM.) But isn't it the same for any of these WMs that if you have errors in your code, your WM can be almost unusable, no? Or is it less likely in haskell?

So far I've never done that hacking on Xmonad; the type-checking means only runtime errors happen, and I'm better at Haskell than I was at Common Lisp, so I have yet to, say, reload Xmonad with an empty root keymap. (With that example in particular, I'd actually have to work hard at it. The usual style for configs is to *combine* the default keymap with the custom keymap - 'myKeys c `M.union` keys default config c'. So if I replaced myKeys = [], I would still be left with the default XMonad keymap when I reloaded.)

(And please make it habit to record the date of last web page edit, which is really informative for readers.)

gwern.net is a Darcs repo, so the full history is available. Or you could check the HTTP headers for the last-modified header, I believe. I've been meaning to ask Jasper to make Hakyll provide the last-modified file time for easy substitution (I'd probably put it below the 'abstract' field), but haven't gotten around to it.

# Wow, you're interested in 柄井川柳 (Karai Senryu)? I don't know # about him at all, but I'm interested in Edo period senryu, too.

Vaguely interested; that was added years ago when I still thought Wikipedia was a great place to work. My enthusiasm flagged after finishing http://en.wikipedia.org/wiki/Fujiwara_no_Teika and I didn't find very many sources about senryu, so...

# I'm Japanese. :)

I figured from the name. In an amusing coincidence, one of my favorite poets is Fujiwara no Teika (see above). -- gwern http://www.gwern.net

2011/9/6 Teika Kazura :

Thanks, Gwern. So your line of interest is similar to mine. :)

I have one question: the page you referred to says "live hacking [in stumpwm] means you can hose your X session"

It's correct, and I have done it several times in Sawfish, like infinite loops. (I've never tried StumpWM.) But isn't it the same for any of these WMs that if you have errors in your code, your WM can be almost unusable, no? Or is it less likely in haskell?

There are definitely some errors xmonad does not prevent: for example if you write layouts which for whatever reason call `sendMessage' or `refresh' it is easy to get an infinite loop. Perhaps that's why very little code in contrib does that. Adam

On Sun, Sep 4, 2011 at 2:08 PM, Gwern Branwen wrote:

As far as Awesome and StumpWM and XMonad go, I have some random notes on pros and cons which never went anywhere: http://www.gwern.net/xmonad-advocacy

I've decided to simply delete my page. Your http://sawfish.wikia.com/wiki/Comparison_of_extensible_window_managers is better, and I haven't worked on it in ages, nor did it ever become very good. Here's the final page source, if anyone cares: --- description: In what ways is the XMonad window manager better than the Awesome WM? ... # XMonad vs. Awesome ## pros 22:51:58 < jrick> in awesome, tabs are provided through a lua library called Tabulous. unfortunately, you can't have a "tab titlebar" at the top, so you don't know what other windows are in that group in awesome, if you break your configuration and reload, nothing loads up in XMonad, it just won't compile and you can continue using the old one awesome config file keeps changing awesome has many dependencies awesome is less reliable than XMonad, which hardly ever crashes awesome is GPL against XMonad's BSD ## cons XMonad users have to setup third-party statusbars like dzen, while awesome has a statusbar built in (this is related to the dependencies) XMonad uses virtual desktop model by default, while awesome uses tags; XMonad has tag extensions, though awesome statusbar apparently can embed Lua applications in it; eg. a (small) clone of Space Invaders. it can also evaluate lua code non-issues awesome uses XCB and XMonad Xlib; any speed advantage is unnoticeable by the user and swamped by just about anything else (like app rendering) XMonad requires learning at least a little haskell syntax; awesome requires Lua knowledge, but this may be more accessible # XMonad vs. StumpWM ## pros live hacking means you can hose your X session minimal type safety - try and pray? XMonad being compiled means easier and more reliable setup than StumpWM - interpreter setup can be tricky XMonad has smaller binaries and uses less memory than a SBCL with StumpWM loaded; this leads to noticeable performance differences on RAM-limited systems ## cons XMonad cannot really do live hacking different paradigms; possible to do static tiling in XMonad, but not nearly as easy as with StumpWM --- xmonad vs awesome http://www.ghosthacking.net/blog/entry/xmonad_vs_awesome/ -- gwern http://www.gwern.net