[lsb-discuss] Don't blame LSB and standards, please: was: Re: Fedora Plasma Product, feedback please

Mats Wichmann mats at wichmann.us
Mon Mar 31 16:44:53 UTC 2014


On 03/31/14 09:34, Jeff Licquia wrote:
> On 03/31/2014 07:06 AM, "Jóhann B. Guðmundsson" wrote:
>> Last time I check and from dawn of time the LSB standard required
>> application to be packaged in RPM format which immediately excludes
>> distributions that do not use RPM as their default/preferred package
>> manager
> 
> This is not correct.  LSB applications may be packaged in a number of
> ways.  One of the possible ways is a RPM format, which supports a subset
> of RPM's features that can be supported on non-RPM systems.  (Basically,
> we only support RPM features that are also supported by the "alien"
> package conversion tool.)

To put this another way, a conforming system is required to be able to
consume (including via conversion) applications packaged in the
LSB-specified RPM subset. There is no requirement that that be the
native package format at all.

>> Why should an application or distribution strive to follow and meet
>> those standards when they are not in the buisness of selling or
>> supporting that distribution, application or application stack since to
>> me that standard has always seemed to be more written to favoring those
>> that make profit out of GNU/Linux ( Red Hat/Novel etc) and related
>> software rather than being focused on standardization/unification in the
>> GNU/Linux ecosystem.
> 
> I'm sorry you have that perception.  We try to be strictly neutral
> regarding distributions, and one of the goals in our most recent efforts
> is explicitly to more aggressively target the community distributions.

I have material on this for multiple blog posts, which I'll maybe write
up some day :)

It's not a surprise that the direction of a project is set by the
interests and contributions of the participants.  For a long time, many
of those who participated in LSB were more motivated to think about an
ecosystem for commercially supported distributions.  You could argue
that more direction was set by companies like IBM and Intel by way of
the people they contributed to the project (disclosure: the latter paid
my way on the project for a decade) than by specific distro providers
but it is undeniable that "enterprise distributions" were more a target.
That's not as true today, as Jeff notes.  There's still always been a
lot of attention paid that the distros the LSB developers actually use
(Fedora, Ubuntu, OpenSUSE, Mandriva, so on) on a daily basis are capable
of conforming to LSB.  Whether it makes sense for such distros to
"certify" is really a separate question, and probably not a terribly
important one.

>> Few bugs open does not mean that standard is well written it might just
>> as well mean nobody is following/using thus have faith in it and the
>> fragmentation in the GNU/Linux ecosystem itself is evidence enough that
>> LSB is failing as standardization body since it does not solve the
>> problems it initially was created to solve.
> 
> Do you have concrete suggestions for how to solve those problems?

We get told periodically that in fact, there's no longer any
fragmentation to speak of, so LSB isn't needed any more to solve that
problem.  That seems rather at odds with your perception, and it would
be interesting to expand on your views a bit more. There are certainly
blips as technology evolves where life gets a little complicated for a
while, but it does seem to me today's Linux systems are remarkably
similar in the core libraries they provide.  If you go higher up the
stack to, say, deskstops (Gnome2, Gnome3 shell, Unity, KDE, XFCE, etc)
there are differences but that's been beyond the scope of a project like
LSB anyway.

Historically, the main issue LSB was built to solve was to have a
consistent (binary) interface for applications, so they didn't have to
be ported to each distro.  Now nearly 15 years later, it looks like on
desktops/servers there really isn't an appreciable market for "shrink
wrap" type binary apps - ones that expect to run on multiple different
distros.  There's a huge market for binary apps on Linux - and it lives
over in the mobile space, primarily Android, which tackles the problem
an entirely different way (there are not "different distros", only
successive dessert-themed releases of Android).  So maybe LSB isn't
solving the problem it was created to solve, but that's more because the
problem isn't one that people seem to be worrying about any more.  Apps
target the one or two distros they really care about, and don't support
any others, and that's about the end of the story. For everyone else...
well, we all run just distro-packaged open source software anyway,
right?  :)

So if the Linux community thinks there's still a fragmentation problem
that a standardization effort could help, people should come forward
with it.  LSB has certainly never been a prescriptive standard, it
counts on contribution and consensus to determine what can be considered
common ground.  Perhaps that in the end has been a failing because it
means things evolve fairly slowly, but a prescriptive standard where a
standards body tells people what they have to do would have been widely
sneered at and ignored - in effect, it would have "failed fast".


-- mats


More information about the lsb-discuss mailing list