Conversation
webusb
webhid
webbluetooth
webserial
webmidi
webgl
webgpu
please make it stop
10
0
0
document formats weren't meant to access arbitrary and complex pieces of hardware
1
0
0
needle
Show content
@v sorry, im not letting a random web page put needles in me neocat_up__w_
0
0
0
re: needle
Show content
@aetios @v wdym?
0
0
0

@novenary why tf did they make webserial when webusb could very much act as a serial connection 😭

1
0
0

@novenary like this is probably a hot take, but something like webusb to allow websites direct access to devices isnt bad
but do we deadass need them split up like that, cant it just be one “webdev” and the browser handles the abstraction

1
0
0
@r0se idk specialized APIs can make sense in this case like, webhid is the same (hid can be a bluetooth or usb device)
for the same reason why the kernel has usb-serial drivers and hidraw
i still object to letting web pages do any of this though
1
0
0
crude and dumb joke
Show content
@novenary

webass

allows website to inspect your ass
1
0
1
re: crude and dumb joke
Show content
@tiredbun this is pretty close to reality unfortunately lol
0
0
0

@novenary oh yeah fair
i think its pretty nice to have a web page do that, like for example the graphene os or pixel web flasher is imo MUCH better than having to use some weird native app that probably brings a ton of security risks with it

1
0
1
@r0se yeah for sure it's pretty neat, but I am extremely conflicted by the state of the web in general and all of these APIs contribute to browsers being bloated monoliths and impossibly difficult to implement
on top of the potential security nightmare because gating it behind a permission prompt is absolutely not enough to protect users who are conditioned to accept everything without reading
1
0
0

@novenary oh yeah absolutely, some proper seperation from these “web” APIs and just websites would be nice to have, though i cant really think of any way to do that in any good way

and the permission prompt thing is like, you can usually get gullible users to run a random executable just as easily, i think device security in general is just fucked

1
0
1
@r0se i kinda like the way firefox has done it for webmidi, where they require a site-specific extension to be installed
apparently they're also neutral about webserial and wouldn't be opposed to a similar scheme there?
they still (rightfully imo) refuse to implement webusb/hid/bluetooth though
i think at least for usb devices some sort of opt-in metadata in the descriptor tables would make it more acceptable, of course restricted to specific origins, kinda like how WCID is an opt in for winusb (the native userspace USB driver)
1
0
0

@novenary how does the site specific extension thing work? is it like a browser extension?

1
0
0

@novenary ohh i see.. interesting
does the site permission link work for you? im getting a 429 error lmao

1
0
0

@novenary lmao it loads in firefox but not on helium
crazy work

1
0
0

@novenary yeah https://helium.computer/
its.. okayish tbh and i actually dont remember why i switched to it but im too lazy to go back to firefox now

1
0
0
@r0se fair lol, personally i can't live without tree-style tab so switching is not an option for me
0
0
0

@mikoto @novenary SCSI, the universal protocol for storage devices

1
0
1
@koakuma @novenary soon to be NVMe once we get spinning rust speaking that protocol (may or may not happen)
1
0
1
@mikoto @koakuma I think nvme might potentially be too tied up with pcie semantics? either way USB MSC is primarily SCSI and UASP is SCSI-only, so it'd have to be ported over there too
although I suppose USB4/thunderbolt should allow it via PCIe tunneling anyway lol
2
0
0
@novenary @koakuma there is nvme over fabrics (NVMeOF), one version of which works over TCP.
1
0
1
@novenary @koakuma so in theory that same scsi over websocket trick that was mentioned earlier could also tunnel nvme commands...
0
0
0
kink
Show content

@xarvos @novenary i'm willing to bet that's well represented on fedi

1
0
0
re: kink
Show content
@ozzelot @xarvos if most definitely is
0
0
0