>> I should like to know why more apps don't require the *bsd
>> {proc,kern}fs interface.
Near as I can figure, it goes like this:
Nobody mounts them because nobody uses them.
Nobody uses them because they're never mounted.
> Under modern BSD4.4, the preferred method is using sysctl(3),(8), as
> opposed to kernfs.
Except kernfs exports some things sysctl doesn't - at least under
NetBSD, the variant with which I'm most familiar.
What are the sysctl analogs of /kern/msgbuf and /kern/rootdev, in
particular? (/kern/rootdev is especially nice because it's directly
mountable; with sysctl you'd first have to create a writeable
filesystem somewhere to make a /dev entry in, leading to a
chicken-and-egg problem.)
Also, the filesystem interface has advantages over the sysctl
interface, for some uses. In particular, the interface to userland is
*much* more stable, meaning that kernel/userland mismatches break it
significantly less often.
der Mouse
[EMAIL PROTECTED]
7D C8 61 52 5D E7 2D 39 4E F1 31 3E E8 B3 27 4B