[ On Wednesday, March 10, 1999 at 15:50:09 (+0100), Soren S. Jorvang wrote: ]
> Subject: Re: RFC: /kern/summary
>
> I think the variables that Brian suggests adding to kernfs
> would fit in well as sysctls.

Just to add some "fuel" to the fire:

I personally think sysctl's API, both from the shell programming point
of view, as well as from the C interface, sucks big time.

There are also very very critical problems with the current
implementation of sysctl on the inside.  I personally think it's a messy
hard-coded hack.  The half-completed attempt to make it more like a true
ASN.1 MIB in FreeBSD (or is it completed now?) was more on the right
track, but it still didn't quite come as clean as it really should have.

I believe it should be possible to give sysctl a filesystem interface,
perhaps within the guise of kernfs, perhaps as a new implementation.
This shouldn't have to be an either/or decision.

> What I'd really, really like to see is a getprocent(2) (or something
> like that) syscall. I imagine it returning a generalized proc structure
> and the pid of the next process in the list. Or perhaps all processes
> in one 

This is way too old-fashioned for me!  ;-)  But it would work much
better than any kmem groveling code ever can.

-- 
							Greg A. Woods

+1 416 218-0098      VE3TCP      <gwoods@acm.org>      <robohack!woods>
Planix, Inc. <woods@planix.com>; Secrets of the Weird <woods@weird.com>