After seeing so many "dump and run" code drops on sourceforge, and now on github
I think the big difference between the two sites is that it's much easier for these types of projects to gain momentum and continue being updated on GitHub.
On SourceForge, a project would linger with no activity, bugs and feature requests would pile up, and creating a new project for a fork was a pain. I've checked out project CVS/SVN trees from SourceForge to fix things and since the projects looked dead, I never contributed the patches anywhere. I certainly wasn't going to create a new SourceForge project just to host my one patch. The few times I've created bugs for the patches, they never got any attention.
With GitHub, it's much easier to fork a repo, fix a bug, and push the fix back up for everyone to see, even if it never gets turned into a pull request or accepted upstream, lingering in your own git repo never to be touched again. Since GitHub shows your fork on the original project's fork tree, it's easy for other users to find your fork with your bugfix and continue that momentum by forking your fork, fixing their own bug, and so on.
But how often have you looked for a program that does something and found a dozen that only do 3/4ths of what you wanted? Just search for open source CD rippers for example, the few good that exist are swamped with results for crappy ones that somebody hacked together in a weekend.
Do people find value in unmaintained codedumps? I registered a dedicated github account last year for all the small stuff I wrote during train rides (I do a lot of them, 5 hours apiece) and never really finished. The idea was that I put a README on each of those and just leave it there for anyone to look at. But most of them are just tiny experiments in Ruby, so I decided not to release them for precisely the reasons mentioned in the parent post.
I really like reading other peoples code (even more when the topic/language/whatever is interesting to me) even if it is only some small proof-of-concept or idea / unfinished code, i.e. not really useable as-is.
I do, sometimes. I've learned from toy/unmaintained/experimental projects. If I'm trying to learn how to effectively use some library or framework that isn't hugely popular, I'll look at any code I can get my hands on, even if it's from some random codedump.
We do this with the Seattle Clojure group. We frequently live-code collaboratively on a small problem that can be tackled in 2 hours. Usually it's silly and useless, but a few have been picked up or at least come in handy to point others to as an example of a certain technique: https://github.com/Seajure
I've certainly found some like that useful. Admittedly for didactic purposes--when I'm learning something new, I like to look at existing code. I've also sometimes found solutions to problems I'm having from random projects which just happened to solve the same problem.
Admittedly it doesn't happen often, but it does, so you may as well put them out there :).
I's like to open source my cancelled game, Shadow Physics, at some point, as well as with a bunch of other small games and prototypes, but I'm also not a fan of "dump and run" code drops. I guess I could pull out the novel and interesting stuff and gut or at least clean up all the cruft and hacks, but once you've already abandoned a project it's hard to find the motivation to clean it up to so other people can learn from it.
What would make the world a better place? More well documented AND maintained examples of code that satisfies some need.