[OpenAFS-port-darwin] Re: 1.2.10 and patches? (Ragnar Sundblad)

[Henry B. Hotz]

At 12:01 PM -0400 8/25/03, port-darwin-request@openafs.org wrote:
>Date: Mon, 25 Aug 2003 12:17:39 +0200
>From: Ragnar Sundblad <ragge@nada.kth.se>
>To: Sebastian Hagedorn <Hagedorn@uni-koeln.de>
>  >> The Mac OS X high-level APIs use stat() to examine access rights for
>>>  files and directories.  It trusts this data, and uses it to determine the
>>>  user interface.  For example, if a directory is mode 0, the Finder will
>>>  show a lock icon on it, and will not let you open it, even though AFS
>>>  will still let you open the directory if you have the proper rights in
>>>  the ACL.
>>
>>  OK, I understand that. But what would be the alternative? How do other
>>  OS's deal with this issue? Should and do they actually try to read or
>>  write a file instead of relying on stat()? Doesn't this incur a severe
>>  performance penalty?
>
>I believe that the common answer to this is that they should
>use access(2) instead. There seems to have been a debate within
>apple some years ago whether there is a performance penalty
>with access, and I think that there is under certain
>circumstances and that is supposed to be why they use stat
>instead. Maybe the problem is that they want to stat anyway
>and they figured that one call is faster than two.
>
>I don't know how this has proceeded. Maybe someone else on this
>list knows more.
>
>/ragge

There was a comment on darwin-development that the relevant call was 
buried somewhere in their OOP framework and they were having trouble 
finding it in order to fix it.

That was a while ago.  If anyone wants to submit (polite) feedback on 
their comment web pages it might help.
-- 
The opinions expressed in this message are mine,
not those of Caltech, JPL, NASA, or the US Government.
Henry.B.Hotz@jpl.nasa.gov, or hbhotz@oxy.edu