Re: Linux Core Consortium
Name changes are a superficial design flaw that obscures the
fundamental design flaw in this proposal -- sharing binaries between
Linux distributions is a bad idea to begin with.
Fixing ABI forks, and articulating best known practices about managing
ABI evolution going forward, that's a good idea. Building an open
source test kit that exercises the shared ABIs, validating that the
test kit builds substantially the same on each distro, and helping
ISVs resolve issues that the test kit missed (and add them as new test
cases), that's even better. But if two competent packagers, working
on different distros, can't get the same ABI out of the same source
code, then upstream's build procedures are badly broken -- and I don't
want that papered over by passing binaries around!
>From the point of view of a user of commercial software, I want to do
business with ISVs that take responsibility for the proper functioning
of their software on a system that is "reasonably compatible" with
their anticipated target environment. Exposed ABIs, as verified by a
test kit, are an appropriate standard of "reasonably compatible".
It's in everyone's interest for that test kit to be correct and
thorough, which is a good thing, because it's a lot of work to build
and maintain it.
I prefer open source platforms for a number of reasons. One is that
it's the source code, not any particular binary, that is authoritative
about how things should work. In principle, one can bug-fix and
rebuild any components in any order without breaking the system. My
experience has been that the more faithful a distro is to this
principle, and the more work is put into abiding by it, the more
likely it is that I will be able to use its binaries unaltered!
That's one reason why I choose Debian when I have the option.
ISVs and IHVs who think that binaries shared among distros will help
them manage tech support costs and quality issues aren't thinking
along these lines, perhaps because testimonials like mine tend to be
anecdotal. Maybe they could be persuaded by quantitative evidence.
I'm not in a great position to gather that evidence; perhaps someone
else is?
Cheers,
- Michael
Reply to: