Subject: Re: conditional symbolic links
To: None <current-users@sun-lamp.cs.berkeley.edu>
From: Ty Sarna <tsarna@endicor.com>
List: current-users
Date: 11/23/1993 06:45:33
In article <199311121641.LAA00539@orchard.medford.ma.us>
sommerfeld@orchard.medford.ma.us (Bill Sommerfeld) writes:
> This worked on Domain/OS because the equivalent of namei() was in user
> space on that system (and thus, it had access to the environment).
> 
> The kernel knows nothing of environment variables except that they're
> a secondary set of parameters passed to a program across exec().

What most people want this kind of feature fo is pathname substitutions
based on machine architecture, like AFS's @sys feature.  That can be
done without looking at the environment, since the kernel already knows
what kind of architecture it's running on.

I'm planning to add @sys to NetBSD unless someone beats me to it.
Hopefully, it can be commited to the official tree. [I say hopefully,
since some people will complain it's a gross hack... Personally, I think
it's very clean compared to the alternatives, such as automounter games,
shadow trees of symbolic links, etc. Unneccesarily complex solutions to
a simple problem, if you ask me. AFS's @sys is simple, it works, and
makes life much earier on people administering heterogenous networks].
I haven't yet decided wether to implement this in lookup() or in the
individual VFS lookup routines of the filesystems that would benefit
from it. I'm leaning towards the latter.

-- 
Ty Sarna             "Don Johnson is back -- back in time, and battling
                      drug-trafficking dinosaurs. Don Johnson -is-
tsarna@endicor.com    Jurassic Narc. Tuesdays at 8; 7 Central & Mountain"

------------------------------------------------------------------------------