Bruce Dubbs <[email protected]> writes:

> However, I think the focus is too narrow.  There still needs to be some
> sort of definition of what it means to be a Linux OS.

Why?  This is a sincere question: I'm curious what problem you have that
you believe would be solved by this definition.

The one that comes to mind personally is API (not ABI) compatibility: can
I compile this piece of software on the system.  But in practice this is
achieved by every distribution using the same basic seet of libraries, or
is solved by API compatibility at the alternative library implementation
level, rather than at the Linux definitional level.

I'm reminded of all the effort that went into defining a POSIX shell (and
I include myself in this; I've spent many hours on the topic and memorized
all sorts of arcane rules required to write portable shell scripts).  Yet
in my experience, outside the rarified world of people who write extremely
portable code for a hobby or are trying to make dash work as /bin/sh in a
distribution, everyone just uses #!/bin/bash and the bash man page and
ignores POSIX.

Observation is not endorsement.  I like standards, and personally would
feel more comfortable if all this behavior were written down in formal
documentation, and there were multiple competing implementations following
the same specification.  But my unease with the chaos of programming to
implementations isn't a very strong argument for the amount of ongoing
engineering effort required to maintain a standard.

-- 
Russ Allbery ([email protected])              <http://www.eyrie.org/~eagle/>
_______________________________________________
lsb-discuss mailing list
[email protected]
https://lists.linuxfoundation.org/mailman/listinfo/lsb-discuss

Reply via email to