Re: DogCows or Polymorphism in the Hurd

[Patrick Negre]

I have read all the thread on DogsCows, and i can't see why it's not possible 
to avoid binding views to new names.

Where the following construction is false or unusable ?

 Hypothesis :
1) Each candidate translator on a file present a coherent sub-filesystem over 
the file.
2) Each file which have multiple candidate translator on it, must have a 
default translator for POSIX calls.

3) Each process have his particular fs-view, and the view depends on which 
translator the process choose over the files it opened.
4) A process can open at most a view over a file, so his specific fs-view is 
non  ambiguous.
5) A process inherits the fs-view of the process that raise him.
6) A process can temporary set the default view over a file, of course under 
rights constraints.

Using this hypothesis avoid the bindings and renamings over a file.
The 2) resolve the POSIX calls problems.
The 5) provide that a called programs will share the same view, and then can 
open the same files that the caller, a property that is wanted, IMO.
The 6), is a result of the crypted directory example, a user who use the 
decryptor translator may want that all process can access the unencrypted 
content of the directory, so the default translator must be change.

Regards Patrick Negre.

        

        
                
___________________________________________________________________________ 
Nouveau : t�l�phonez moins cher avec Yahoo! Messenger ! D�couvez les tarifs 
exceptionnels pour appeler la France et l'international.
T�l�chargez sur http://fr.messenger.yahoo.com