I never bothered to understand pool difficulty before.  When I started out with GPUs last year, the default suggested difficulty on my account/worker page had always been 1.  When I switched to ASIC last month, it stayed at 1 because I didn't know any better but in spite of this, BFGMiner indicated diff at 13.  Reading the recent posts on here prompted me to edit my worker just now and picked 17-32 GHash/s range.   My account/worker page now shows a suggested difficulty of 16 and BFGMiner diff is also 16.  What exactly happened here and what is the significance?  Is this the right/optimum setting?

Edit:  BFGMiner now shows diff went down to 10 (I'm confused).  BTW, since I made the changes, my Rejects went up form .08% to .21%, if that is even significant.

If the difficulty is bigger than 1 you have less transmissions and you produce less traffic. Given the payment method and taking into consideration that some rounds last less then 1 Minute, or only few minutes, there may be a problem running on higher diff.  With all my USB cockroaches running at diff 1 I observe higher rewards in short rounds (considering similar Overall hash rates, and running 24/7). My explaination for this used to be different Internet Connections (mine is 50 Mbit/s) or higher diffs. Of course I dont complain.

From my experience, it makes no difference whether you have 100 workers each with 1 device, or 100 devices all on 1 worker.  It should all work out the same at the end of a round.

One question regarding split to more workers:
Is it possible to split workers and lower difficulty if I'm using stratum proxy on one IP address?
Link bandwidth isn't problem for me.

EDIT: mining with cgminer on Linux - running multiple processes = workers isn't problem

I've tried to create new worker and split mining processes but difficulty is adjusted at stratum proxy.
So if two workers go outside thru one stratum proxy difficulty goes up.

Conclusion: I need to use more proxies for more workers to keep lower diff

When will Slush pool implement Namecoin in the mining pool? I left Bitminter and I miss the namecoin which was also mined at the same time.

There's even nothing like work at Diff1. :-)
They just work. They calculate hashes. Most of the hashes have values that are higher than the one necessary for even Difficulty 1. Only one out of every four billion hashes (statistically, in long run) is small enough to reach "difficulty one" and potentially become a share.

And thanks for your patience, Trasla.

Why does higher difficulty mean higher rejects? (Does it really?)

Thanks for the explanation.

Did he say why?  Just curious.