On Sun, Jul 09, 2006 at 02:38:24AM +0200, Lars Marowsky-Bree wrote: > On 2006-07-07T23:39:16, Alan Robertson <[EMAIL PROTECTED]> wrote: > > > I think this is getting FAR FAR too complicated given the probability of > > it being useful. And, it isn't obvious that the most important kinds of > > semantics are captured in the discussions... > > Partially. Attaching permissions to each object in the CIB, treating it > somewhat like a unix filesystem hierarchy, allows all you described > below, _and_ is a way of doing it generally, in a way which doesn't > require the CIB to understand itself. > > But, you sort of describe my original proposition, which was to attach > permissions not to specific objects, but to specific operations on some > of these. > > What you describe cannot be done in the cib process itself, not sanely - > these operations are known to the mgmtd only, the CIB just sees > create/deletes/updates/queries on specific elements and doesn't > understand their meaning. > > I came from the approach you seem to take, but I think I've turned > around, since Dejan's suggestion seems to be more general, and can > express more, and seems to fit in better with the general design of the > CIB. > > One thing which you do describe, which _is_ missing from the Dejan model > (I shall call it that for now ;-) is that it only expresses permissions > on existing objects (but then, very powerfully so). It does NOT express > which objects may be created (and with what permissions). That would > seem to be required.
I don't think it's missing. Given that, for example, a user has a write permission on the <resources> section, they can create resources. If they have write permission on a particular resource, then they can create attributes for that resource. Something like when UNIX user has write permissions on the directory then he can add/remove files from that directory. Cheers, Dejan > Otherwise, a work-around would be to pre-create them, and allow them > write permission to the attribute which makes the object effective or > not; that again would work, and seems fairly obvious to use too. > > > Sincerely, > Lars Marowsky-Br?e > > -- > High Availability & Clustering > SUSE Labs, Research and Development > SUSE LINUX Products GmbH - A Novell Business -- Charles Darwin > "Ignorance more frequently begets confidence than does knowledge" > _______________________________________________________ Linux-HA-Dev: Linux-HA-Dev@lists.linux-ha.org http://lists.linux-ha.org/mailman/listinfo/linux-ha-dev Home Page: http://linux-ha.org/
