Linked by Thom Holwerda on Thu 8th Nov 2012 20:54 UTC, submitted by Elv13
Thread beginning with comment 542043
To view parent comment, click here.
To read all comments associated with this story, please click here.
To view parent comment, click here.
To read all comments associated with this story, please click here.



Member since:
2006-03-01
you know that a kernel has one ABI facing userspace AND one facing drivers (or in general, modules), right?
and that OTOH userspace applications are all linked against the standard c library which actually implements syscalls - thus effectively decoupled from the kernel - but there's no decoupling layer between drivers and the kernel?
that on the userspace side, keeping the kernel ABI stable is actually unnecessary (as long as the library is updated accordingly when disrupting changes are made) and it's really enough to maintain compatibility at the library level - but it'd be necessary (and a sign of good sw design and development practice - ie compartmentization) on the kernel side (which ironically is the one ever changing)?
now, linux actually got it backwards in this regard (and also wrt desktop IPC...)...
and although users can adapt to the current situation for them to choose between using a certain kernel version OR supporting a certain piece of hw, and risk breakage/regressions (not that there arent any, see aspm..) updating the whole kernel, is a suboptimal situation
De Icaza may have been too far reaching in that sentence, but he wasnt completely wrong...