Going back to older software

For twenty-three years, the window manager for my desktop was a venerable piece of software called 'olvwm'. It was very fast and responsive, and had mostly very good ergonomics.

But olvwm represents very much the Path Not Taken in graphical UI development:

• It supports Focus-Follows-Mouse wholeheartedly, albeit this is not even the default
• The physical screen is taken to be a mere window onto a wider expanse, at least four times as large; you can pan around this virtual screen, and dynamically increase and decrease its size
• There's a monitor for the virtual screen, showing tiny rectangles transparently representing all your programmes' windows, and you can manipulate them directly via this monitor
• The whole thing is agnostic about whether you want windows to overlap ...
• The tooling for iconising / deiconising windows, and raising and lowering them in the z-stack, is extremely snappy
• If you don't want windows overlapping, there are also affordances for selecting groups of them and moving them about en bloc. I don't think I've seen this in other window managers.

But olvwm is a also good example of software decline. It hasn't really been developed since about 1993. It depends on several key libraries that themselves have stopped being developed. This eventually led to it only really surviving via redstributors, such as Fedora, Debian and NetBSD. Bit by bit, the code lost support: the underlying libraries weren't ported forward for 64-bit operation, 32-bit emulation stopped being an option, and eventually the dependencies and then olvwm itself ended up being dropped by redistributors. Ultimately even the source code became a bit sketchy to find.

I had to stop using olvwm in 2018 due to an Ubuntu upgrade. I switched to fvwm, but it remains sluggish by comparison. Yesterday, however, I just happened to think that AI might make it easier to fix the 64-bit portability issues and so went Googling. Needless to say, there appears to have been quite a few people who've tried to preserve or maintain olvwm and its ecosystem over the years, and in 2021, someone made a fresh start on Github, putting up a good copy of the most recent source code, and archive links to where to find other relevant material.

By random coincidence, around the same time as I started looking for the code, the new maintainer started working on it again. By the time I'd got it compiling on my own machine, he'd even pushed a new commit or two. I was able to switch my main desktop over to olwvm for the first time in eight years; it'll take a few more tweaks before it can be my daily driver again, but it was wonderful to have it back - it was as always impressively fast. It was odd to see how quickly the muscle memory came back, too.

One of the positive effects of AI has been a big discount on technical debt: it won't be too hard to get olvwm and its dependencies back into production. It's very nice to go back to a stable desktop system. I had the same config file from 1995 to 2018; there was no need to change my behaviour every few years to jump through a hoop for some product manager I'd never heard of. The software was finished and Just Worked.

Gemlog index

Site index