[lsb-discuss] [Lsb-infrastructure] LSB 3.1 Update 1 status (and minutesfor2/20 and 2/27)

Wichmann, Mats D mats.d.wichmann at intel.com
Fri Mar 2 09:21:44 PST 2007

Vladimir Rubanov wrote:
> Could anyone please describe what optional LSB module mean in terms of
> certification? Well, it is clear that we have specification
> and tests for optional modules. It is also clear that if an 
> implementation does not support functionality from an optional 
> module it is OK for certification. But what if it does support 
> the functionality from an optional module? Do we require that 
> it should behave as in LSB in this case (i.e. do we require
> passing the tests for optional modules for those distributions
> that support them)? What is the sense in writing specification 
> and creating tests for optional modules in case we do not 
> require anything about them?

you've put your finger on what has always been the issue
with anything "optional" in LSB - what does it mean for

Ian's crafted the current message there but essentially
it's best to look at optional modules as something which
clearly *will* meet our Best Practice criteria but does
not yet, usually because it's too early or out of cycle
for a few.  Case in point is indeed Qt4 - targeted at
LSB 3.1 which is supposed to run on RHEL5 and SLES10,
we were then told that RHEL5 wouldn't have the support.
This made Qt4 an optional module, to become mandatory
later.  Providing this specification early in the LSB
allows people to get a look at it, and possibly even use
it. It also allows us to get the work in and published
while the contributor is interested - if we'd waited
until today to start Qt4 work I'm not sure we would have
access to the resource that did the work as he was 
foolish enough to get a promotion within TrollTech!
Ian likes to refer to this concept as "late binding" -
we do work as if a feature would go in, and at the last
possible moment decide whether or not to throw the
switch which marks it "required".

There is no requirement on distributions; however,
we'd *like* them to say the also support optional module
FOO and applications using FOO would then run on that
distribution.  In the dtk-manager context, it would
actually have been easier if Qt4 was a separate test
suite rather than a runtime flag to the desktop suite.

More information about the lsb-discuss mailing list