[OpenAFS] Re: directories in afs have "owners"?

[Derrick J Brashear]

On Thu, 22 Dec 2005, Adam Megacz wrote:

>
> Derrick J Brashear <shadow@dementia.org> writes:
>> And it's directories, not files, for which the implicit ownership
>> being talked about comes into play.
>
> Okay.  Would it be correct to say that the ownership of non-directory
> files matters only for the "extra permission check" ("u+rw" bits are
> checked in addition to AFS ACLs)?  Or are there other situations where
> this information gets used?

mode bits/unix owner are only ever checked by the client itself. Server 
doesn't care. (modulo directories as mentioned)

> Does the "group" of a directory (or file) in AFS space have any
> meaning?

No.

> Also, since PTS userids don't have to match local UNIX userids, do UNIX
> applications get confused when they create a file and fstat()->uid !=
> getuid()?  Or just "no more confused than usual when dealing with AFS
> file semantics" (ie only very rarely)?

Most sites do match uid to pts id, but, most applications don't stat() 
what they create anyway.

> Lastly, is there any reason for files to have ownership other than
> preventing dumb users from clobbering their own home directory's
> permissions?  It seems like there would have been some other way to
> accomplish this with a smaller "semantic footprint"...

In AFS, you mean? Because AFS was supposed to be like the ufs, presumably.

Derrick