libera/#shirakumo - IRC Chatlog
Search
12:47:49
Colleen
<shinmera> 5s google shows this https://github.com/Mashweb/docker-project-for-radiance
12:50:49
Colleen
<shinmera> As far as I understand docker there shouldn't be much that needs to be done out of the ordinary. Should work the same as hosting any other web app.
12:52:42
shka
for instance, you make build step, it is fine, you can save-lisp-and-die tough because a thread appeared
12:53:46
shka
so instead I write another entrypoint file, but when i try to load radiance it starts to complain that my package does not exist
12:55:55
Colleen
<shinmera> you can sbcl --eval '(asdf:make :radiance)' to get a base deployment structure, put your own module into its module dir, and that should work, iirc.
20:52:09
Colleen
<shinmera> It seemed to claim to have some sort of static typing system, but I couldn't really figure out more than that and some of the caveats seem very bizarre.
20:58:28
Colleen
<shinmera> right. I'm much more interested in leveraging type inference and templating, myself.
20:58:41
Colleen
<shinmera> I did have a prototype of such a thing for my 3d-vectors rewrite, but I never finished it.
21:09:03
Colleen
<shinmera> Like, I'm not interested in anything that doesn't integrate with anything else. Why would I use that? I could just go and write Haskell or whatever else that's already proven to be good for that.
21:10:05
drakonis
now, the only thing that baphomet has that interests me is the bit about language oriented programming, but i can get that through ruricolist's vernacular
21:10:45
Colleen
<shinmera> I haven't looked at coalton yet, but my gut feeling is that it's a step too far.
21:23:43
drakonis
never let it be said that CL can't be the all paradigm language with the right libraries
21:25:57
Colleen
<shinmera> Overlord is another thing I meant to look at but dropped out of my head.
21:27:16
Colleen
<shinmera> State tracking cannot be as hard as I was making it out to be last time I was working on it.
21:31:46
Colleen
<shinmera> the main point is that as long as we can separate the thing orchestrating the build from the thing performing the build we can avoid poisoning, which is all we want, really.
21:32:58
Colleen
<shinmera> being fully functional in the build is also nice, but a headache too far for me :)
21:38:12
Colleen
<shinmera> Well anyway, Forge has an outstanding issue that prevents it from running a full-scale test, namely I couldn't figure out how to properly track build progress across builds, to eliminate repeating of steps that don't need doing. It's messy due to the distributed nature. Though I think there should be a system that isn't overtly complicated, and I just need to think about it with a clear head again.
21:39:38
Colleen
<shinmera> typically build systems simply rely on file timestamps to figure out what needs redoing, since each build step is pure in that side effects only appear on the file system
21:39:57
Colleen
<shinmera> however, a build target like a lisp image has side effects directly in that client
21:40:12
Colleen
<shinmera> and we can't retrieve when those changes were made from the client itself
21:41:26
Colleen
<shinmera> further complicating things is the question of hosts and logical addressing of build data
21:41:55
Colleen
<shinmera> I'm too tired to coherently explain all the issues, and would need to refresh myself from the code base anyhow.
21:43:58
Colleen
<shinmera> but it is the last part I can tell that's standing in the way of using Forge to attempt to build a real system.
21:44:30
Colleen
<shinmera> Which would then show whether the algorithm I devised to compute the action plan is even scalable or not.