LWN: Comments on "Kernel Summit 2006: The user-space API"

Kernel Summit 2006: The user-space API

nix — Wed, 19 Jul 2006 09:35:04 +0000

The different platform versions are actually in subdirectories, not ifdeffed out. HAL's not *that* ugly. :)

Kernel Summit 2006: The user-space API

dlang — Tue, 18 Jul 2006 16:28:29 +0000

the HAL api presented to userspace should be cross platform, but the part of HAL that does the work is very platform specific. a cross platform version is just all the specific versions combined with ifdefs and similar

Kernel Summit 2006: The user-space API

dlang — Tue, 18 Jul 2006 16:27:00 +0000

when something is in the kernel it gets more maintinance becouse when people change an interface they useually go through the kernel and change everything that uses that interface to match (and there's a good argument that doing so is (or should be) required by whoever wants the change)

this by itself goes a long way towards solving maintinance issues. it doesn't cover all of them (especially the fact that sometimes changes shouldn't be just propagated, sometimes they should trigger much more significant changes), but it also significantly reduces the frustration of whoever is maintaining the system so that they just work on real fixes and improvements rather then chaseing updates.

Kernel Summit 2006: The user-space API

cventers — Tue, 18 Jul 2006 12:28:19 +0000

It's interesting and surprising that they would consider adopting Hal. I
think Hal's a very necessary project and needs some work, and bringing it
into the kernel source would probably help. But it sounds like a little
bit of a crazy idea. Isn't Hal supposed to be multiplatform?

Perhaps what is needed is just better communication between Hal and LKML?

Kernel Summit 2006: The user-space API

jospoortvliet — Tue, 18 Jul 2006 12:27:15 +0000

HAL = 'unmaintained crap'. wow. didn't see that one coming...

Now I wonder - why would bringing something into the kernel tree ensure better maintainance?

And would bringing as much as possible into the kernel not make things harder to maintain, as I can imagine its a bit harder to start contributing to the (big) kernel compared to a much smaller project?

And what's the reason to push as much as possible to userspace? Yes, you can upgrade it, and it can be maintained by others than the kernel dev's, and you can fork it. But udev and hal are such 'outsourced' pieces of code, and now the kernel hackers think about making them part of the kernel (tough still userspace). So its not just positive, isn't it?