On Sat, 25 Mar 2000, H. Peter Anvin wrote:

> Philip Rackus wrote:
> > 
> > I personally believe that layering the LSB is a **good thing**, however 
> > (there is
> > always a however)  the trap is that if you create to many layers, the 
> > possible
> > combinations of layers means that you don't have a standard at all, and 
> > your back
> > to square one.  Layering can work if it is kept to two or three layers at 
> > most.
> > 
> 
> I think POSIX did it very well: create a set of options.  Then define
> option groups as so-called profiles.  To be compliant with a specific
> profile you have to support all the options that constitute that
> profile.
> 
> However, I would be *much* happier putting this discussion off until we
> have one published and preferrably accepted standard.  Time is running
> out, quickly.
> 
>       -hpa
> 

The fact that time is running is a good arguemnt for splitting the LSB
into layers so that you can find acceptance for a smaller part first and
you have a much better defined goal to work towared. When layer 1 is
defined, move on to the next layer and so on. 

//Mats Loman

Reply via email to