At Thu, 13 Oct 2005 14:42:26 +0200, Bas Wijnen <[EMAIL PROTECTED]> wrote: > > I already concluded that, but we need them anyway. We want to be mostly > POSIX-compliant, and that isn't possible without them.
This is true, but with a catch: We want to provide a POSIX personality. This does not mean that we need user IDs at the lower levels of the system. Rather, that user IDs can be provided in addition, on top, or orthogonal to the actual system. One thing that was always a goal of the Hurd project was to have multiple personalities. But currently, this is not the case, and there is no real foundation to do so. The Hurd servers are in some aspects quite POSIXish. Not always conceptually. For example, the Hurd authors got the signal handling mostly right. But the actual details of implementation are tightly build around POSIX. What Jonathan offers us here is a system which is at the low level very secure and robust, and some perspective on how to build native interfaces and application components within that system. The POSIX personality can be added on top of that, or orthogonal if you prefer. I have given it not enough thought yet, but there are basically two models you can think of. You can either have a user-local POSIX "shell" which connects the POSIX emulation with the rest of the system. Because this shell is user-local, it would be completely owned by that user. Or (but not exclusive-or), you can have a system-wide POSIX "box", which connects the native system to the POSIX system inside, plus it provides a way for users to connect to the POSIX system via a terminal if they have the necessary capabilities (they could get them from the system administrator, for example). Within that POSIX box, your "user ID capability" would be worth whatever it is worth in the POSIX world, but outside of the POSIX box it would be worthless and authenticate you for nothing. This are only rough ideas yet, but I hope it is clear how this would work. Thanks, Marcus _______________________________________________ L4-hurd mailing list [email protected] http://lists.gnu.org/mailman/listinfo/l4-hurd
