Notices by Josh Simmons (dotstdy@mastodon.social)

@hyc @nikitonsky if people never step outside their comfort level they never find out whether something actually is hard or not, and they have to rely on the equally uninformed heresay that floats around.

@hyc @nikitonsky I would perhaps put this differently, that people often have wildly inaccurate estimations of complexity. In both directions (for different cohorts). So it's silly to suggest you'd just bang out a quick caching system and have it be robust in production, at least without a lot of work. But it's also silly to suggest that X person is incapable of it. If they take the problem seriously they're going to be just as capable as anybody else, more or less.

@whitequark 9V games are the ones you can lick

@whitequark one file per provider is the only true distributed solution. Then you'll only be missing one part of your software if a vendor goes down.

@whitequark @wingo and yeah i never really expected that it was imposed (im certainly not a 'redhat truther'), i just find it super interesting how groups of people ingrain ideas collectively, and also then how difficult it is to dissuade those groups from harmful ideas. (and i mean, all groups have some of these unfortunate stuck ideas too, it's not exclusive to gnome)

@wingo @whitequark right I just found that also, it's linked from the old gnome hig https://wiki.gnome.org/Design(2f)HIG(2f)Planning(2f)Configuration.html the power of One Guy

@whitequark yeah i'm just curious, was it like a stated aspect of apple's design that everyone copied? or did somebody make a decree about it back in the day. I don't really remember when the 'zero configuration' push started. Is it all because everyone hated xorg.conf so much :') anyway it's really funny coming from a games background where you'll get lampooned (rightly!) for not supporting configurable fov, motion blur, configurable bindings, 100 graphics settings, accessibility modes, etc.

@whitequark i'd love to trace where the linux desktop userspace world got the idea that configuration was bad. it's such a weird ingrained thing across many projects.

@bob @whitequark @gsuberland @untitaker yeah, that would make sense, but you can't oversubscribe a static document store. at least not without doing something impressively wrong.

@whitequark @gsuberland @untitaker I have to imagine it's some kind of backoff? E.g. it's quite literally doing nothing. Because from the screenshot this is also a static document???

@whitequark @gsuberland @untitaker WEB SCALE

@whitequark @gsuberland @untitaker now i want do know what they're even doing for one and a half seconds...

@whitequark @0xabad1dea did they fuck up their company name and they meant Thanatos to match with Prometheus? xd

My partner is learning python (zero programming experience) and wanted some introductory resources, are there any good ones out there? to save me the effort of wading through the endless sea of SEO optimised AI garbage...

@skinnylatte suburbia enjoyers D:

@whitequark @0xabad1dea @gsuberland @tthbaltazar another thing which does behave that way in practice is GC and systems built around GC shaped data-structures. e.g. if you trigger an emergency collection on oom or low memory. which, exactly like you say, doesn't work so good when you have multiple instances. (classic 'language server' issue there, manually pressing the force GC button in resharper to salvage memory...)

@whitequark @0xabad1dea @gsuberland @tthbaltazar Honestly I'm not sure it's common that applications speculatively fill up memory as cache. It's true mostly just for things like games which dynamically adjust their texture budget for GPU memory, but even there lots of them just splat out whatever they require and hope vidmm keeps their shit resident when they're in the foreground. cpu side applications seem even less sophisticated in general, but still end up relying on OS paging.

@whitequark @xgranade webgpu itself is also not really sufficient as a platform api, too many missing pieces.

@whitequark incredible, now that's the client-side rendering i can get behind.

@whitequark @joe @charlotte @8051enthusiast it's interesting in rust because you only need the atomic because of sync issues, otherwise a single threaded atomic is exactly a cell, I believe. I'm not sure whether going towards "atomic" is better from an API perspective, or if making "cell" more ergonomic is a better choice.