paul@DELRIO.CC.UMICH.EDU ('da Kingfish) (06/08/88)
The problem is that whatever verifies passwords doesn't distinguish between not getting a valid password because the registries are not reachable match or because it was interrupted. The same thing happens when you crp. If you interrupt it an the right moment, it will use the local registry instead of the master because it failed to successfully query the master. We had a situation where Joe Luser changed his passwd but was still getting whacked by his putative friends who would get to use his old passwd from a local rgy that hadn't been updated this way. I think you get logged in as user.none.none because user.none.none has no password, no? We assigned a password to user.none.none. Control C is interrupt, so you check the local regy and it comes back ok because there isn't a passwd. if there *is* a password, and you still get logged in, then there is something a might peculiar going on. So, I called in some of the APOLLO people that I knew, and the response I got back was this: "Well, it is basically a feature that APOLLO has included so that if by any chance you mess-up your registries and there is no way that you can get in to your system, then you can get logged in as user.none and try to recover your registries". I was told the solution to this problem is to delete the user.none.none entry from our registries and then the problem should be fixed. The feature part is getting user.none.none if the registries are unavailable. The bug part is the system thinking they are unavailable because the user interrupts it. As far as removing user.none.none goes, that seems pretty rad. As I mentioned, we just gave it a password. --paul