[plug] Named pipe. Solution to programming problem?

[James Devenish]

In message <apb072xgvq.ln2 at shameless.plug.org.au>
on Fri, Nov 19, 2004 at 01:09:29PM +0800, Mike Holland wrote:
> James Devenish wrote:
> >>isn't that what a named pipe does?
> >Not in a way that satisfies all the requirements I infer from Tim.
> >But maybe his requirements are simpler than I imagined.
> I don't get it. Whats wrong with a named pipe here?

If you know how to prevent race conditions, broken pipes, etc, then
you know more than me. Tim indicated there were multiple sources
supplying the data and, for the lack of contradiction, I assumed there
are multiple readers (or, at least, there are unknown random intervals
between read attempts). At the very least, Tim seemed to indicate that
the readers could not be reprogrammed with a locking mechanism on top
of which I assumed that the inability of FIFOs to cope with multiple
readers was inherent in the design. Hence, I did not think that named
pipes could satisfy Tim's requirements. But I may have been mistaken
through inexperience or past misadventure.