Re: FreeBSD devfs support on NetBSD 5.0

[Iain Hibbert <plunky%rya-online.net@localhost>]

On Fri, 19 Jun 2009, Alistair Crooks wrote:

> On Thu, Jun 18, 2009 at 11:22:59PM +0300, Antti Kantee wrote:
> > On Wed Jun 17 2009 at 16:26:42 +0200, Adam Hamsik wrote:
> > >
> > > On Jun,Wednesday 17 2009, at 4:16 PM, Hubert Feyrer wrote:
> > >
> > > >On Wed, 17 Jun 2009, Reinoud Zandijk wrote:
> > > >>I dunno, i am not for `a' working implementation, i'm for a *good*
> > > >>implementation. I must agree that Matt's version is taking its
> > > >>time, but
> > > >>hopefully he'll finish it soon.

I've heard it said that waiting for perfection means that you will wait
forever. Having said that, I agree with Reinoud in that I would prefer a
good implementation over just something that works. It is common enough
that when something works nobody will bother to make anything better and I
like to think that here at NetBSD we strive for good.

Matt writes:
>       "If HPS has a decent, working implementation of devfs that
>       gets you the same features that I was proposing, then I'm all
>       for integrating his and scrapping mine.

HPS does not have such a beast at this time from my reading of the thread,
he has an older version running on NetBSD-4.0 and was wondering if it
would be worth it to have the hooks in the tree so that his upgrade would
be easier in future. I didn't see any specific proposal.

So,

- is there a design document for the FreeBSD devfs?  I looked at devfs(5),
  devfs.rules(5), devfs.conf(5) and devfs(8) but didn't see an overview

- was FreeBSD devfs considered and rejected in the past and if so what
  were the reasons?

I did have a conversation with Matt some months ago about the status of
his devfs and the situation at that time did not meet my definition of
good as there was a structural magic that I disliked. The FreeBSD devfs
does not completely work for me either (at first glance, why is 'at boot
time' rules devfs.conf different to 'at run time' rules devfs.rules? and
why aren't the rules specified at mount time?)

I think that tying it all up into a single massive 'add devfs' change
event could be a mistake.  Primary need is a notification method in kernel
that can signal when devices are added or removed. The whole dynamic /dev
thing can hang off that and I'd be happy in the first instance with a
daemon that created and removed nodes as required with mknod/unlink..

iain