You can do this with delegates. you have to code safe, to avoid errors when
there are no subscribers (unless you add a "default" subscriber like a log).
To have it so outside software can hook in, you will need to build a
lightweight publisher/subscriber type of model. I do not have links right
now for these, but a google on delegates (or better yet, multicast
delegates) and publisher/subscriber (not sure how to frame that one) should
yield some patterns and sample code, if nothing else.
--
Gregory A. Beamer
MVP, MCP: +I, SE, SD, DBA
*************************************************
| Think outside the box!
|
*************************************************
[quoted text, click to view] "Chris Johnson" <cej@nightwolf.ogre.uk> wrote in message
news:%23zZDyxGgIHA.4744@TK2MSFTNGP06.phx.gbl...
> Hi all,
>
> Just a quick question as I'm not entirely certain what phrase to google
> for as all attempts so far have failed to illicit anything useful, and the
> best anaology I can think of comes from hardware :-)
>
> I'm curious as to whether there's some form of API in .NET that's the
> software equivalent of diagnostic pins on a circuit board? Essentially I'm
> want to have a bucket I can throw diagnostic message into in a
> fire-and-forget method, and not care whether there's something looking at
> the bucket (connected to the pin) or not. If nothing's listening, the
> message gets forgotten about.
>
> Ideally it needs to be lightweight with little latency -- the intention is
> to use the 'pin' as a performance diagnostic, so in normal situations,
> nothing would be connected. If there are problems, then we can attach to
> the pin and read timestamp messages. An event queue would be no good:
> queue would eventually fill-up and hang the code, unless there was extra
> logic in the code to clean the queue up, which could start getting messy.
> Performance counters won't really work either as that then requires a
> large chunk of logic to keep track of individual entities in order to
> create an "average latency" counter.
>
> Essentially it's analogous to attaching an oscilloscope to a circuit board
> and looking at the latency of stuff between two pins on a chip, etc.
>
> I suppose shared memory might be one possible solution, but I understand
> there's no CLR wrappers for shared memory (excluding third party
> extensions). If anyone's got ideas or pointers, I'd be grateful to hear
> them.
>
> Cheers,
>
> Chris
>
>
>
