Re: What is "real-time"?

[munnari.oz.au!manuel.anu.edu.au!coombs!gsc@uunet.uu.net (Sean Case)]

hansen@micro.cs.umn.edu (David M. Hansen) writes:

>A better definition, I think (this is not original BTW), is that a
>real-time system must _manage_ time as a resource.

This is the definition I've been working towards for some time.
Do you have a reference for it?

>Certainly, it must meet all its deadlines (I imagine that word has
>an interesting etymology), but also it must delay taking action on
>a calculation if that calculation completes before its answer is
>required.

More generally, it must perform calculations taking into account the
time elapsed since its inputs were valid and the time that will elapse
before its outputs are used.

Consider a radar tracking system: it has a position for an object at
some point in the past, and must calculate where to look for it the
next time around.

Now, one way to improve management of any resource is to use less,
and speed does help. But a fast system that thinks it's infinitely
fast is often less useful than a somewhat slower one that properly
accounts for its own usage of time. Ideally, we'd like a fast system
with proper time management, but this seems to be beyond the
capabilities of most contractors...

(Off the top of my head, a deadline was originally a line around a
prison beyond which the guards could shoot to kill. I think.)

Sean Case
--
Sean Case               gsc@coombs.anu.edu.au
"[...] if a poststructuralist doesn't get you,
 a deconstructionist will."--Ursula K. LeGuin