Terribly, terribly sorry for the recent list problems. I thought we'd eliminated the bug but something ate the entire userlist today. This really bothers me... I hate having problems and I take personal pride in how well the Phoenyx usually runs. Can't do this kind of thing for over 13 years without doing it right. We're really working on this problem as best we can, but Karen's in her fourth month of her first pregnancy and just recently got over severe all-day sickness (why do they call it "morning" sickness, anyway?) But she still tires easily and tends to go to bed an hour or two earlier than she used to. And since she's the programmer, I don't dink with the code much, especially since Karen doesn't believe in documenting for other people to read it. For those who are interested in details about the problem... There two kinds of userlists... the "user database" that keeps all the interesting info about everyone... alternate addresses, "real name", group settings, etc. Then there are the distribution lists, which are the delivery addresses filtered out of the database for easy manipulation during delivery. There's one for the main list, one for digest, and then one for every subtopic. We lost most of the main database. Since the distribution lists were still there, and that's what's used to actually deliver mail, mail that made it through to the list went out to everyone. But approval to post to the list comes out of the database, so most of you didn't have permission to post. (This is where the confusing "you are not subscribed" comes from when you're still receiving mail. As far as the list robot is concerned, you *aren't* subscribed... even though you're on the list to get mail. :) We *thought* we'd found the bug... there was a point where a process was neglecting to lock the user database file before reading it, which meant it could read while another process was writing it. No error occured during the writing, but the reading process would read an incomplete file, do it's job, and write that incomplete file back out. Karen's working on figuring out why we're still seeing this kind of error after fixing that bug. -- Carl D Cravens (raven@phoenyx.net) Why get even, when you can get odd?