path name rules

[John (J5) Palmieri]

On Mon, 2005-10-03 at 18:15 +0200, Marcel Holtmann wrote:
> Hi Havoc,

> > 
> > If we allow all characters, it will be a big pain in the butt since
> > paths will have to be escaped wherever they are _used_ (in html, shell
> > scripts, language bindings, etc.) So it's a whitelist of characters and
> > paths have to be escaped when they are _created_.
> 
> using characters with no need to escape is a good idea. However
> crippling the character set is bad, because the conversion can also
> become a big pain.
> 
> > Why underscore and not : is basically because it doesn't seem worth
> > going through ascii and thinking about each character one-by-one, and it
> > seems easier for dbus programmers to remember "just alphanumeric and
> > underscore" than "alphanumeric and random list of exceptions"
> 
> What is about '-'? This will make IEEE addresses more readable then
> using an underscore or non separator at all.

There was a patch added that enabled '-' for bus names.  This has caused
some problems and I am almost tempted to revert it if it doesn't get
fixed or there was no compelling reason for adding it.  I am hesitant at
this stage in the game to add any spec changing code unless it is really
warranted.  Object paths are just implementation details and having them
be human readable especially when dynamically generating them is not all
that important.  More important is getting this stuff to 1.0.

-- 
John (J5) Palmieri <johnp at redhat.com>