[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: X and LSB

Michael Stone wrote:
>> Now I'm confused. DQ is saying that LSB is trying to find common ground
> between the distributions in order to give ISVs some assurance that
> their apps will run on some large portion of linux systems. How can you
> accuse him of attempting to *impose* standards? IIRC, you're the one
> suggesting that current practices be abandoned, not him.

No, I'm discussing a point.  But I'll respell out what I think should
happen again, because I think it's time to cut through the discussion
and restate my position on X again.
Separating off an X layer should be done in the LSB.

If the layer is separate, it can then include more.  If it includes
more, ISVs who need X will be more satisfied with the result.

LSB will "test" conditions, not impose required packages (in my
understanding), so if it tests for X, let it test for X, not just a
subset of X (a few libs).

Call it whatever you want, a layer, or a level.

But to require part of X to satisfy some ISV requests is not going far
enough to help them significantly.  An ISV that needs X will still have
to find a way to test for X at that point.

And, by subset of X, the LSB is going a bit far in the countless
situations that X is not needed.  Database servers, routers, and
countless applications from embedded to massive number crunchers do not
require X.

Simply specifying more X components be tested for will help the ISVs who
need X.  Separating those X components into a separate layer will help
the LSB apply to a much broader base.

This is nothing new.  This isn't even my idea.  This was something that
was discussed over a year ago, and was accepted as a good idea by many
at the time.  And, one of the greatest comments I remember on the LSB is

"To me at least the final success of the LSB will be measured on how
much it manages to avoid specifying rather than how much it manages to
specify."  -- Alan Cox. 

I think that if X is going to be specified, it should be done in a way
that standard software that uses X can count on all the base components
working as expected.  This wouldn't be done by mandating or endorsing
any specific X software.  This would be done by testing that the
functionality of the common X components are there using a test suite. 
And I am fine with that, let there be X.

But, it should be done so that the base itself can stand free of X.  As
a fully functional X layer on top of the base.

Reply to: