[OpenAFS-devel] Solaris afs.rc file damage

[Dean Anderson]

On Thu, 26 Apr 2007, Sean O'Malley wrote:

> On Thu, 19 Apr 2007, Derrick J Brashear wrote:
> 
> > We use the workaround code we have for that eventuality.
> 
> I apologize but I get scared when people start talking about workaround
> code.. =)
> 
> How good is the workaround code? Can we get similar performance out of it?

That's the pseudo driver code we were talking about. I don't think we
know the comparision. Its running on Linux, and its 'similar' by some
standard--we just don't know what standard ;-) 

I'm convinced that people are rightly concerned, and that the syscall is
preferred and will be faster, if by an unknown amount.

> I am wondering whether it would be smarter to just plan to slowly dump all
> syscalls now and incorporate it as a design feature rather then wait.

I think a common framework that makes the choice most nearly invisible
to the rest of code is a good approach, and then wait and see what Tom
Keiser's code does and what the difference in performance is. The
current code may be so slow by comparision as to make the difference
unnoticeable at present.

But I'll be pretty happy if we can remove the reboots from the rc
scripts, and do more controlled reboots only during installation. And no 
reboots at all are preferred.

> I do realize it takes time to do correctly. Although since it is mostly
> IBM licensed code they might have an "easy button" for it. *ducks* :)

 ;-)

		--Dean


-- 
Av8 Internet   Prepared to pay a premium for better service?
www.av8.net         faster, more reliable, better service
617 344 9000