Subject: Re: Slow output from 'ps'
To: None <perry@imsi.com>
From: Chris G Demetriou <Chris_G_Demetriou@lagavulin.pdl.cs.cmu.edu>
List: current-users
Date: 07/02/1995 04:32:55
> Chris G Demetriou writes:
> > I have several comments on this:
> [...]
> > 	(2) making procfs mandatory adds bloat to the kernel.
> 
> [ ... ]
>
> BTW, kernfs, procfs, the sysctl stuff, and /dev/kmem crawling all seem
> to overlap on the functionality Venn diagram. Unix usually provides
> one good way to do something, and we seem to be accumulating lots of
> cute ways to do things that aren't being used.

you seem to have ignored my first (?) point, which was that we would need
to keep around the "/dev/kmem crawling" in any case.  (if you'd like
to know what, 'man ps'.)

as you've noted, sysctl() overlaps memory-crawling significantly.
However, it provides functionality that memory-crawling cannot
provide, and it's sufficiently inflexible that it can't be used in
place of /dev/{,k}mem.

kernfs and procfs as meant to be examples, in my opinion.  same goes
for the 'fdesc,' 'portal,' and 'umap' file systems.  Yes, all of them
are potentially useful (and, as i understand it, some people have
actually found them useful), but to my mind they are primarily there
as examples of how to do 'odd' things in file systems.  I would argue
that there is merit in having things around solely as examples.



chris