I'm also still getting large numbers of stales. About 6.5% on average, most of them right before Long Poll work pushes.
What miner are you using? The reason that I ask is that Phoenix 1.48 was giving me far more rejected shares than poclbm (latest). I suspect this is because how many more shares there are in a round, but I am not sure why that should make any real difference.
Phoenix 1.48 with phatk. It worked perfectly fine before the DDOS event, and it's only this pool that has the problem. I'll try poclbm in a bit, though. Thanks for the advice.
I think BTCMine is probably a bit slower with long polling push due to the rapid expansion that occurred when deepbit.net went down from the DDOS and so many of its users flooded into BTCMine [and didn't leave
after deepbit came back up]. poclbm should help, but it probably won't be the magic panacea that you are looking for ... maybe it will though; I would certainly like to see the results (I hope that you indicate that over say, 3000 shares, you see a reduction of 50% rejected). dbitcoin indicated that registration was closed until the pool was moved to a new server and I assume that has not taken place yet.
There is only one catch. You will not be able to tell how many are rejected if you use the command line version unless you pipe the output and capture the information [or use GUIMiner v2011-05-21 assuming you are on Windows].
And then my pipe dream bubble below that isn't exactly related:
If I get the time, perhaps I will write a little .NET application to gather all information about pool (address, port, maybe some JSON stats), worker, miner configuration, hardware stats (i.e. video card [with miner], CPU, usage stats, GPU temp ....), submitted shares, rejected shares, date and time of each submission, the hash value itself, long polling pushes, idle time waiting for work (including date time and duration) and of course communication failures. That may not be all that "storage" friendly however, so I may have to decide whether to keep a window of data [configurable?], what type of database to put it in [I think MySQL would work, since the app could then run on Mono under X11 and be happy]. It probably wouldn't be an easy to install application for the average user if I wrote it as I indicate, however, reducing the storage and and using a light weight local DB could work as well, but personally, I would love to see information for months or years at a time just out of curiosity. I don't think putting the back end hosted on the Internet would be cost effective though since most people wouldn't use it and almost certainly wouldn't pay monthly for it. I'll write it for myself at some point and maybe a few people will have interest in it
I wish I had more time to do it in the near term, I have a lot going on in my life right now [Daughter graduating high school and heading off to college, overtime with work, yada yada].