(Illustration by Gaich Muramatsu)
Hi Patrick, On Wed, Apr 13, 2005 at 10:55:57AM -0600, Patrick Walsh wrote: > # cfs la /coda/myrealm > System:AnyUser rl > System:Administrators rlidwka > # ctokens > Tokens held by the Cache Manager for root: > @myrealm > Not Authenticated > # ls /coda/myrealm > ls: /coda/myrealm/: Permission denied > > If I authenticate as a user, I can view everything properly. So it it's one of side effects of "to-be-fixed" controversial behaviour when an uid's authentication state changes (say your tokens expire or you do cunlog or you clog as another Coda identity). Your uid has most certainly been used as an authenticated user, and Venus does not handle "properly" the change to unauthenticated. I see it as a useful feature, as otherwise you might be possibly unconsciously using anonymous insecure access when you believe you have tokens (and hence are protected against server spoofing). It will be fixed, but for now let you test anonymous access from another uid, which you never do clog from. It is anyway not a good idea to use the same uid on a host for different identities in the same realm - that way you mix security domains. Coda identities are totally isolated from each other - but all rights on a Unix client are associated with an uid, and as such would overlap both Coda identities. It is hard (if possible at all) to handle such situations and keep the access semantics intact. Note that you can have tokens for different identities in _different_ realms, at the same time, it is of course totally ok - it causes no semantical problems, the uid's rights are the plain sum of the rights in the realms. In contrast, it is undefined how to "sum" rights of two identities in the same realm, there is no support for such operation. Each action on objects in Coda is done with one sole identity. In unix you want a separate uid for each identity to make sure they do not (possibly badly) interact. I do not say your particular case can not work - it could - given some fixes and also some constraints which are possibly not trivial. Cheers, -- IvanReceived on 2005-04-13 14:16:11