>>>>> "Stuart" == Stuart Hughes <[EMAIL PROTECTED]> writes:
Stuart> Support for priority inheritance is an officially sanctioned
Stuart> mode for POSIX mutexes. This is an important mechanism that
Stuart> can save a system from complete lockup. I agree that one
Stuart> should not design a system to explicitly rely on priority
Stuart> inheritance for its correct operation, but if you have a
Stuart> complex system which you cannot completely analyze and test
Stuart> 100%, this may be the one thing that saves your system from
Stuart> entering a fatal state.
Um, that sounds a bit inconsistent. Let me paraphrase what I heard,
tell me if I got it wrong and why...
1. You shouldn't design a system that relies on priority inheritance
1a. That's because it doesn't necessarily help? or
1b. That's because you can't tell whether it will help or hurt? or
1b. That's because it can turn a working system into a broken one?
2. Nevertheless, if you design systems complex enough that you can't
analyze them, you might as well throw this in because it may -- by
good fortune -- happen to unscrew an otherwise screwed up situation.
(Or it may not. Or it may make it worse??)
Somehow, the notion of building a system where you don't actually know
if it is built right, and putting in some things that aren't
necessarily a good idea on the theory that maybe it'll fix things,
isn't exactly comforting. I know Microsoft builds apps that way, but
is this a way to build control systems?
paul
-- [rtl] ---
To unsubscribe:
echo "unsubscribe rtl" | mail [EMAIL PROTECTED] OR
echo "unsubscribe rtl <Your_email>" | mail [EMAIL PROTECTED]
---
For more information on Real-Time Linux see:
http://www.rtlinux.org/rtlinux/
