On Mon, 23 Mar 2009 09:27:48 +0100 "Stephan A. Rickauer" <stephan.ricka...@startek.ch> wrote:
> On Sat, 2009-03-21 at 12:14 +0100, Henning Brauer wrote: > > * jmc <j...@cosmicnetworks.net> [2009-03-11 15:05]: > > > so anyway, how are _you_ using probability? > > > > it's high on my list of useless features in pf I'd rather remove. > > if anybody is actually using it, I'd like to hear about it. > > Once in a while a re-spot this 'feature' in the man pages and find it > very cool. But then I can't come up with any idea of how to use it > sanely. Could that be a case of 'uselessness'? ;) > > (never had to simulate bad lines so far, have enough of real ones) > Using pf's 'probability' feature to simulate bad lines is creative and *might* be useful for the most simplistic types of testing, but it is really the wrong way to preform a this type of testing. The right way to simulate bad lines is to use (expensive) equipment designed to inject various types of errors/conditions into the line in known and measured way (www.spirent.com), and then use more (expensive) equipment to detect the errors on the other end (TTC/Acterna/JDSU). Anything less, and you're only guessing. With probability, there is a remote chance you'll either inject no errors at all, or conversely, only inject errors. It's not a good chance, but it's still a chance, so you really don't know what you're testing, and worse, there's no way to repeat your results. -- J.C. Roberts
