Re: DogCows or Polymorphism in the Hurd

[Marcus Brinkmann]

At Wed, 08 Feb 2006 20:42:01 +0100,
Tom Bachmann <address@hidden> wrote:
> 
> Marcus Brinkmann wrote:
> > At Tue, 07 Feb 2006 15:21:15 +0100,
> > Tom Bachmann <address@hidden> wrote:
> > 
> >>like a directory? Or how could even new apps exchange file names without 
> >>abiguity?
> > 
> > 
> > - Doctor, it hurts if I do this.
> > - Then don't do it.
> > 
> > New applications are better off exchanging capabilities to files
> > rather than file names for access to files.
> > 
> 
> Clearly. I explicitely stated "file names" to point the problem out. 

But new apps are not limited to file names.  The problem here is that
you arbitrarily limit the problem space to exclude the right solutions :)

> Still it might be possible to completely avoid exchanging names other 
> than caps, but if the user is involved this at least requires a way to 
> name a capability by a string. Anyway, this is offtopic.

Ok.  So the new problem is exchanging capabilities using a persistent
directory hierarchy and string data.  This is actually a much more
reasonably restriction than just a file name.  Two possibilities:

1) The application creates a new name for the specific object (facet) it
   wants to bind, and passes a string to that.

2) If no new name shall be created, the application can pass two
   strings: name + desired facet.  Then the receiving application can
   use the poly class members to get the desired facet of the object
   accessed using the name,

These solutions are not equivalent, and which is preferrable depends
on other boundary conditions of the problem situation.

Btw, I don't think this is off-topic at all.  But it would be useful
to more carefully motivate hypothetical restricted situations, so that
we can see what alternatives may exist.

Thanks,
Marcus