"I ran out of fingers and toes to count on."
10 fingers allows you to count to 1,024 (2^10). If you count toes as well then its 1,048,576
You'd have thought such a kernel orientated person would understand binary.
Penguinistas, take heed. The kernel of your beloved OS has rung in the new year with a brand spanking new version number because... Linus felt like it. Linux fans will be relieved to know that while 2019 should feature a gentler, softer and less sweary Torvalds, the man's ability to make arbitrary decisions remains …
You can't count to 1024 with ten fingers. Two fists are 0, all fingers stretched on both hands is 1023. I could think of one way to show the 1024, but that is only possible for male persons and you shouldn't do that in public (...and you won't come back). Also be careful with 4, 128 and 132, those can easily be misunderstood.
"Why the hell is support for individual hardware a kernel thing?"
I don't think it is. I think you can happily build a Linux kernel with most of that hardware support chopped out and then load the drivers you want separately as modules, but the normal development practice is to include all those drivers in the kernel-space source tree (so that they can agree on interfaces) and so they get released together.
It's normal development practice because Linux kernel doesn't have an ABI. Changes to the kernel source can and do break modules including drivers.
The easiest way to mitigate that is to ship the module / driver source code as part of the kernel package so that when you build the one, you build the other and they remain compatible.
It suddenly doesn't seem so crazy to suggest that HTML5 support can be included in this process so that we can have some confidence that an update to a web browser, or other program using video, has a consistent, mostly working, place to look. As it is, there are a whole bunch of different media codecs, all supposed to be doing the same thing, except when they don't.
"they didn't have to build so much driver shit straight into the kernel they wouldn't need to do so many releases..."
Sure, and the latest Windows version is 5 better than this. I'll glee in glory at my Windows 10 as soon as my machine applies the 12 update releases since yesterday.
The kernel's version numbering has always been more of an arbitrary counter rather than providing useful information about the scope of changes. It's ungood, and unfortunately is something that has become much more common across the industry.
I want meaningful version numbers!
> I want meaningful version numbers!
Certainly. Consider a version number of the form A.B.C.D
A: Increment this number when you release an update that intentionally breaks existing workflows.
B: Increment this number on a slow day.
C: Increment this number when you release an update that does not break any existing workflows.
D. Increment this number when you release a hotfix for an update that unintentionally breaks existing workflows. (Optional: add a "_rc2" suffix if you can't be arsed to test the hotfix before releasing)
I hope Linux 6.0 is planned more in advance and has all the old and disused interfaces ripped out along with library junk like strfry. It would probably mean 2 years of parallel development to pass the scream test as legacy systems are identified. And this is important since removing the junk will reduce the attack surface and also make the kernel smaller, very useful for embedded work. Over time it will also mean less maintenance work and less work to port to new architectures.
I notice Redox-OS aims for some Linux compatibility but without problems in dropping interfaces they do not like, a luxury they have when the OS is not much deployed.
Biting the hand that feeds IT © 1998–2019