[lsb-discuss] Thoughts on LSB ISO standard

Theodore Ts'o tytso at mit.edu
Thu Jul 26 22:45:04 UTC 2012


So I can't speak for the Linux Foundation, but I was CTO (on loan from
IBM) during the time period when we made the decision to let the PAS
status lapse.  Essentially, the corporate sponsors were no longer
interested; what originaly drove their interest was that there were
various governmental customers who for whatever silly reason, required
an ISO standard "stamp of approval" to get past their purchasing
regulations.  With the widespread success of Linux as a de-facto
standard in the server world, that issue is mostly gone at this point.
So that pretty much took care of the "benefits" side of the equation
from the perspective of the ultimate source of the cash funds
necessary to drive the ISO standards work.

On the "costs" side of the equation, just simply getting the document
into the right shape so that it met all of the ISO requirements would
have required a lot of editorial work.  The people who have the skills
to do that editorial work were at the time, fairly few and far
between, and they all wanted to paid significant sums of money per
hour to do that work.  (Hey, they were consultants with a relatively
rare skill set; I don't begrudge that, but it was still a cost.)

In addition, merely engaging with ISO would have taken a lot of time,
going to various ANSI/INCITS meetings beforehand, and needing to play
politics with (for example) the Sun representative, who had no love of
Linux, and who was distressed that Linux wasn't fully compliant with
POSIX.  There were other all sorts of other silly things; such as
strategizing at the ISO level to have the meetings in the Far East, to
prevent certain known troublemakers from being able to attend (because
they might derail the proceedings) because the travel expenses would
be too expensive for them --- but that also meant that those of us who
had to actually *attend* the ISO meeting had to pay $$$ to travel long
distnaces to far off lands (and if you were employed in by company
with niggardly travel policies, you either suffered in economy class
or upgraded to business class on your own nickle).

All of this pre-meeting politicking and meetings was to make sure
there wouldn't be people trying to make the LSB standard diverge from
reality --- which is one of those things that is always a risk when
you engage with a standards bodies.  Witness what happened with OATH
2.0[1] (which got horribly complicated and declared a failure by one of
the original OATH 1.0 developers) or the XHTML 2.0 vs. HTML 5 fiasco.

[1] http://hueniverse.com/2012/07/oauth-2-0-and-the-road-to-hell/

Basically, after living through the original LSB at ISO process, and
weighing the costs ands benefits, it left me feeling extremely cynical
about the whole ISO process, and all of this was before the OOXML
fiasco.  After all, TCP/IP was hugely success dispite the attempts by
the proponents of the ISO-blassed OSI networking standard.  (And OSI
had all of the benefits of purchasing requirements by silly goverment
agencies; look how much good it did it.)

So for those people who want to push for getting the ISO imprimatur
--- I'd advise you think very carefully about why you want it, and
what benefits you think it will bring.

Also consider the very strong negatives that could occur if people in
the standards world decide that they want to "fix" or "improve" the
standard, like the standard busybodies tried to do with XHTML 2.0 ---
it can turn into a monstrous timesync just to prevent the standards
goeers from doing active damage to a technology.  If there is a
divergence between an international standard and real-life
implementation sanity, the Linux world has very clearly not been shy
to give the big middle finger to the standard (cf the history of
POSIXLY_CORRECT and POSIX_ME_HARDER).

After all, df still prints its output in units of kilobytes by default
despite the POSIX requirement of 512 byte sectors, and if you look at
the success of Linux versus the legacy Unix systems, the blatent
violation of the International Standard hasn't hurt Linux or slowed
its adoption.  Heck, AIX and Solaris ended up adding a Linux
compatibility layer; they changed to meet Linux, not the other way
around.

						- Ted


More information about the lsb-discuss mailing list