sensors framework continued (architecture)

[Alexander Leidinger]

Quoting "Poul-Henning Kamp" <phk at phk.freebsd.dk> (Fri, 16 Nov 2007 20:17:11 +0000):

> In message <20071116205918.0e9d5819 at deskjail>, Alexander Leidinger writes:

> Neither am I denying anybody any voice or chance to argue, but after
> a certain amount of, IMO pretty pointless, argument, I may tire of
> them.

You proposed something in the kernel, I pointed out how it can be done
in the userland (which fits to your requirement of doing as much as
possible in the userland instead of the kernel), and you haven't pointed
out where my arguments are wrong. And I asked you to prove them wrong.
Several times. But you just cut those parts out on your replies without
answering any related questions I wrote. I agree that the discussion
_between both of us_ is pretty pointless and that we have to find other
discussion participants which interact better regarding this. Again: I
try to find people which are either able to explain to me what you
refuse to tell me (and then I'm quiet and you can proceed with your
architectural proposal), or which agree that your proposal is more
complex in the kernel without a significant benefit.

> If you go with the above architecture, you will have answered most,
> if not all critism that has been leveraged against the half-baked
> code imported from OpenBSD, and we can move right ahead to start
> to get things done.

Your comments reads as you have looked at the architecture of the
OpenBSD code. Is this the case, or do you still refuse to look at the
OpenBSD code like you told us several times before?

And in the architecture thread I don't talk about the OpenBSD code. I
talk about using sysctl's to get the sensors data out of the kernel,
and for example Robert Watson also likes the sysctl approach better
than a fd based approach (Robert, if you changed your mind, it would be
nice if you could explain me what changed your mind, as I haven't seen
an argument from Poul which made it obvious for me that his more complex
approach is better than the less complex sysctl approach). Jhb didn't
see anything obviously wrong with the sysctl approach to export the
data to the userland in his initial mail here on arch@ and he didn't
participated in the discussion to explain me where I'm wrong. This means
either he has no time to explain me in other words why your approach
is better, doesn't care how the data is exported, or doesn't want to
get into the line of fire. Those which proposed a fd approach in the
beginning also didn't participated in our ping-pong mails to provide
answers to the questions I asked. So _nobody_ explains what is wrong
with my arguments against the fd approach, even when I ask to point out
things I may have overlooked or get wrong.

Yes, the OpenBSD code is an implementation which complies with the
architectural proposal I talk about here, and if enough people are more
interested in the sysctl approach than the fd approach, it would be
logical to review the implementation and to decide if we adopt it
(which would make our approach compatible with the one of other BSDs),
or if we want to do a different implementation of such an architecture.

Bye,
Alexander.

-- 
Somewhere in Tenafly, New Jersey, a chiropractor is viewing "Leave it
to Beaver"!
http://www.Leidinger.net  Alexander @ Leidinger.net: PGP ID = B0063FE7
http://www.FreeBSD.org     netchild @ FreeBSD.org  : PGP ID = 72077137