glyph

glyph, 11 days ago

@mcc just say “私は強要されているので同意できません。” out loud and tap agree

glyph, 11 days ago

@mcc (Less humorously, I do think that systemd is Good, Actually and I am generally happier to deal with its bullshit than, say, start-stop-daemon, but wow did they ever try to snatch defeat from the jaws of victory in every. single. part. of its user interface)

glyph, 11 days ago

@tef that is an interesting perspective and based on vibes it vaguely tracks, but I'm super curious if you have any specific, concrete examples

glyph, 11 days ago

@tef that is… fascinating. so the kernel devs think that… what… should be in the kernel exactly? UUIDs? PCI device tree paths?

glyph, 11 days ago

@tef oh yeah I am familiar with this genre of criticism, I usually refer to them as "systemd truthers" and I have less than zero patience for them

glyph, 11 days ago

@mcc yeah, my experience exactly. I'm pretty sure I've used it less than you have, but my general impression is that it's like Wayland. They absolutely needed a certain subset of its functionality to get past a modernization milestone and actually boot on current hardware, there were hard constraints about tolerating the old/bad way, but once they got past that they ran out of resources for finishing the thing. That, and it really is badly designed in some ways (I mean, .ini files? come on.)

glyph, 11 days ago

@tef glad to see the resounding success of the design philosophy of (checks notes) the X Windows System continues to drive excellence in the linux ecosystem

glyph, 11 days ago

@tef I mean I'm posting this using Display PDF™ so some alternatives did succeed, but in succeeding they receded into invisibility

glyph, 11 days ago

@tef the magic of the internet is that it makes it possible to find your people

glyph, 11 days ago

@tef the whole mechanism/policy thing is really a bit of a distraction anyway, the problem that this design philosophy runs up against is almost entirely that it fails to define the interface between system services and application programs, assuming that a "distribution" fuses them together into a block of solid metal for you so all the integration concerns get shoved into the user's face. you could have a "pluggable policy" design that fixed this and it would be great.

glyph, 11 days ago

@tef i.e. "mechanism not policy" is a fine interface for the system software infra team to present to the system software product team, but exposing that seam as a public API to the entire world is the mistake

glyph, 11 days ago

@tef oh now you are trying to get me in my feelings

glyph, 11 days ago

@tef but yeah thank you for explaining, this makes explicit several vague intuitions that I've had with how systemd came to operate the way that it does

glyph, 11 days ago

@tef I tend to see this as a pervasive manifestation of "apolitical" programmer brain

specifically, the correct solution to the problem you described is to have a political interface to driver vendors where they have to e.g. register a DNS name with the kernel in order to be able to distribute binary modules that work, a-la CFBundleIdentifier, so nobody stomps on each others' namespaces

but that feels like a busybody telling you what code you can write, so nobody wants to do it

glyph, 11 days ago

@tef well that and everybody still seems jazzed about short, conflict-magnet names like "/bin/fjfj" "/proc/sys/vnfj/blfh0" instead of wanting to have fully-qualified globally unique identifiers anywhere.

case in point I tried typing short line-noise for those examples and had to re-try like 6 times to avoid typing actual names that exist on a linux system I have, that doesn't even have a GUI attached

glyph, 11 days ago

@tef my favorite genre of tech history is "this thing you dismiss as bad for aesthetic reasons is actually very good for solid technical reasons and you should learn why", and name conflicts are a rich source of that genre.

glyph, 11 days ago

@xgranade @tef this is one reason I love talking about Smalltalk, and Squeak specifically. You can run it, it's real! You can point to stuff in it and get ideas from it! And it is so incomprehensibly alien that there is no risk that people will start thinking about it and running into its actual practical deficiencies, so it exists as a pure piece of performance art.

glyph, 11 days ago

@tef @xgranade you can still put a prolog program into a text file and the compile and execute it, which strikes me as a much higher risk of it achieving replication competence and escaping the lab

glyph, 11 days ago

@tef @xgranade it’s true, roy fielding’s thesis was about web browsers, though

glyph, 11 days ago

@luis_in_brief @mcc and my X

glyph, 11 days ago

Hmm. This was the first of a 3-part thread I posted with Ice Cubes and the second two parts seem to have vanished with no trace. Also, threads seem to be limited to 6 posts. So much for a mac Mastodon client that has good threading support.

glyph, 11 days ago

@matt "sometimes anyway" is a load-bearing qualification ;).

I think with the ARM transition the crumbling façade of this design approach is going to finally collapse, and we'll see some trusty old applications which haven't been updated in decades like this finally stop working and need to be updated, or at least need to be run via explicit emulation. But perhaps more than anything else on my list it had an extremely good run.