layered specification

[dumas patrice]

Hi,
(please forgive for the poor english)
I have see all the discussion about the layered vs the broad specification
and I think that there are good arguments pro and cons.

A single specification assures that everyone knows what is lsb compliance.

A layered specification permits to handle special cases for example little
distributions, or distribution that want to customize high level (such as
x windows or, in the future, desktop or window manager level). In my
opinion, there are other interesting things  in that approach, it is
conceptually better and permits a layered work (although I think it is
allready done in practice).

I think there is a simple way to conciliate the two approaches :
there would be a LSB compliance, with the broader scope, and layer
compliance (like core-lsb, X-lsb, window manager-lsb...), compliance with
a layer assuming that compliance with lower layer is met.

This doesn't rule out the confusion problem, but allows two different
approaches in doing the specification : 

top-down : you can specify the lsb (highest layer) compliance first in
order to avoid infinite discussion on where a layer stop, and when this
work is done define the layers in order to deal with the special cases
concerned with this issue.
 (personal opinion : I think that it would fit better in the
official line of lsb).

bottom-up : you can divide the work to produce low layer specification
more rapidely (don't know if it's a real argument as the core could be the
most difficult thing to specify) and build on top of existing layer.
(other personal opinion : I think the division of the tasks in the lsb
teams are appriximately in accordance with the layered model but as I am 
only a mailing list observer I don't really know).


Pat