X and LSB

[tytso at mit.edu]

Let's take 3 steps back and take a look at this whole discussion at a
much higher level.  

What's the problem statement that caused us to start the LSB in the
first place?  Well, here's my take on the situation.

The reality is that many Linux companies are trying to sell Linux boxes
and Linux distributions into various corporate markets, and have been
trying to make further inroads into the desktop market, which is
currently dominated by Windows.  The companies that are doing this
definitely part of the Linux community, but of course they don't
represent all of the Linux community.  Still, I think a lot of people
would agree that what they are trying to do is a Good Thing (tm).

It's also a reality that in order to do this, it is important to get the
support of independent software vendors (ISV's).  They want to be able
to provide binaries that work on "Linux".  They don't want to test
against N different possible distributions; they only want to test
against one interface.  So if we don't provide an LSB, then they will
likely start shipping products based on the dominant distribution, which
today is Red Hat.

Now, how important is this depends on how important you perceive the
impact of commercial software to be.  The fact is that today, there are
certain products for which there are no good free software analogues.
Oracle is one such example.  Another good example is tax preparation
software like TurboTax.  And whether or not you believe that there will
be good alternatives available in the future, many people are not
satisfied with the free software available, and will want to use the
commercial alternatives.  (Personally, I will rather pay $19 bucks to
get Turbotax rather than to spend my own time figuring out my taxes by
hand, or pay for H&R block to do my taxes for me.  So sue me for using
commercial software.)

If the ISV's don't have something like the LSB, they will simply code to
a defacto dominant distribution, such as Red Hat instead.  This is not
healthy for the Linux community, and Red Hat to their credit have been
assisting in making a distribution nutral standard which ISV's can code
against.   However, we need soemthing now.  Some ISV's have already, or
on the brink of, giving up on LSB and simply using RedHat as the
standard distribution.   

Hence, wasting huge amounts of time reopening old discussions about
whether or not distributions should be putting stuff in /usr/bin or in
/usr/local is really not helpful.  In fact, to the extent that it
distracts us from finishing the LSB, it's actively detrimental.

I know that there are a number of Slackware users who don't like the way
most of the other Linux distributions have chosen to do things.  But
that's where the center of mass is today, and we need to standardize on
something that's roughly near the center of mass.  Otherwise the ISV's
will do the job for us, and simply chose one particular distribution,
and it will almost certainly be Red Hat.  If that happens, most other
distributions will need to follow Red Hat, since not being able to run
applications like Oracle will be a significant disadvantage in a large
number of markets where many of these Linux distribution will be
competing.  

This is not the only reason to be pursueing the LSB, but to me it's a
pretty compelling reason.  This is also why I think the embedded space
should be lower priority, since ISV's are much less likely to be
shipping products that have to work in the embedded space.  We need an
LSB that works well for people on servers who want to install packages
such as Oracle, and for people on desktops that want to install Quake
III.  And we need something that does this yesterday.

							- Ted