Alan Coopersmith wrote:
> If we are going to endorse a distro as our reference or blessed or whatever
> distro, what are the requirements it has to meet?
>
> Here's a list of ones I can think of to start discussing from:
>
> 1) 100% Open Source: The OpenSolaris Constitution, as approved by the
> voting members of the community and the Solaris management at Sun,
> requires:
>
> All software produced by the OpenSolaris Community shall be
> licensed to the public free of charge under one or more open
> source licenses approved by the Open Source Initiative.
>
By that token, Indiana can never meet that requirement. There are
closed source components in it, that are not likely to *ever* be opened
up, and for which it is not realistic to find a suitable open source
replacement.
Removal of those closed source bits (or replacement with open source)
will cripple the distribution on certain types of hardware, in some
cases unacceptably so. (I'm talking about hardware such as LSI
SCSAI/SAS/SATA controllers, Nvidia graphics chips, etc.)
Don't take this as a belief on my part that this requirement is not
appropriate, nor as a belief that there is not substantial value in what
Indiana offers. I just don't think *what* Indiana offers is precisely
what we want from an OpenSolaris reference. (And, to be quite honest,
because of the above limitations, I think a reference distribution is
likely to be of somewhat limited interest by end-users.)
> 2) Decisions about the distro will all be made by an OpenSolaris community
> group in accordance with the constitution (which can be oversimplified
> down to "just do it" for simple/obvious things, "quick e-mail consensus"
> when the answer isn't so clear, "formal vote" for the important things.
> See Article VIII for the full details).
>
Ok. Such a group needs to be identified and chartered. I don't think
the Desktop group is the right place for that, and (last time I checked)
Indiana isn't a CG.
> 3) All components architecturally reviewed in the open by the process and
> groups established by the OpenSolaris Architecture Process and Tools
> community.
>
Agree.
> 4) Supports the platforms designated as Core Platforms by a community
> process TBD (initially SPARC 4u/4v & x86/x64).
>
Agree. We may need to identify "reference platform implementations",
though. (Because of the hardware driver problem with open source
already mentioned.) Notably, no platform with an Nvidia or recent ATI
graaphics chip is likely to be acceptable. At the moment, recent
hardware that this *could* be done for looks like only certain SPARC
low- to mid-range SPARC hardware and x86 systems based around Intel
chipsets.
> At the moment, I don't know of any distro that meets all of these, not even
> Indiana, though it and a couple others may be able to achieve them with a bit
> of work.
>
I don't think any of the current distributions have the *goal* of
meeting these criteria. Frankly, as I've already noted, the *reference*
distribution is likely to be less useful on so many different hardware
configurations that it really shouldn't be what user's download by default.
If OpenSource is truly the goal of the reference, then it really needs
to *not* be plastered all over the main os.o web page. Its much more
likely to be interesting only to hardware aficionados, distribution
developers, and maybe certain ISVs looking to be able to test for core
compatibility.
> Software that didn't meet those rules (closed source binaries, NDA covering
> ARC review) could still be installed on the distro, but couldn't be a core
> part of it.
>
That's a different idea. But it sort of defeats the purpose of having a
community-defined reference, I think.
I really, really think the community needs to decide what it wants out
of a "reference distribution" before deciding whether any one
distribution qualifies.
I'm of two minds of the value of a "reference" distribution. What is
most needed, is a binary downloadable distribution. Frankly, naming
confusion aside, SXCE did achieve most of that. And, at the end of the
day, it will be whatever Sun ships (regardless of names) that will be
the measuring stick that ISVs and users judge other distributions
against -- *regardless* of what the community decides.
At the end of the day, I think for end-users, ISVs, and the community,
it is best if Sun is able to ship a binary product called "Sun
OpenSolaris". If there is a different edition, perhaps called
"Community OpenSolaris", or "Nexenta OpenSolaris" (or whatever), that is
fine. But "Sun OpenSolaris" will be what the vast majority of folks
download, use, and judge compatibility by.
I can imagine having something like LSB from Linux, maybe "OpenSolaris
Certified Core", or somesuch, which identifies core components, to help
ISVs. But I predict that such a thing would take a long time to get any
traction in the community.
-- Garrett
