William Kucharski wrote:
>
> On Mar 8, 2008, at 4:06 PM, Martin Bochnig wrote:
>
>> If such a CG is needed, then this name makes the most sense.
>
> I prefer "Community-Supported Platforms" to "New Platforms" as it makes
> clear that:
My "New Platforms" was just raised as food for thought, but a few points
below.
>
> 1) The ports will largely be performed by the Community, not directly
> by Sun engineers (though Sun may aid in providing reference code or
> reference ports for certain platforms.)
Why do you think this is so? Isn't it possible that some Sun Labs
project to port Solaris to an interesting piece of embedded MIPS
hardware could take place? Isn't Sun just another community member here?
>
> 2) The ports will be maintained by the Community, not directly by Sun
> engineers (though Sun may aid in maintaining certain reference
> ports.)
Again, why are you making a distinction here? Who *cares* who maintains
the port? If Sun blesses the port by issuing "official support" for it
(e.g. making a regular Sun Solaris release for it), that should not be
any different than say, Nexenta doing the same.
>
> 3) Someday some platforms will not be "new." For example, after the
> PowerPC port is completed and maintained for a number of years, does
> it still belong in a Community called "New Platforms?"
No, it doesn't belong there. Once a port has reached a level of
stability and is integrated, it probably can either fork off a new
project or community of its own. Any port with an active user base and
vibrant community (one out of incubation), deserves to have a first
class community. I don't think either the PPC or S/390 ports are at
that stage yet, but the x86/amd64 and sparc ports definitely are.
>
> Don't get me wrong, changes made to OpenSolaris to enhance the
> portability
> of common code will make it back into the ON tree, but processor-specific
> code for ARM, PowerPC, etc. will NOT go back into Solaris as a product.
When did you start making those decisions -- do you have a crystal
ball? I can easily see Sun releasing PowerPC or (more likely) ARM or
MIPS code getting into Sun released products (perhaps not as banded
Solaris product, but maybe... more likely as a part of some closed
product, such as a network appliance of some sort.) One can imagine
Sun wooing some of the larger vendors (Cisco? Bay Networks?) to
consider Solaris for some future network infrastructure project, which
might well use an unconventional architecture... in which case an
official "Solaris" port might make more sense.
I see no reason why PPC, ARM, or MIPS couldn't get integrated into the
ON consolidation, provided it is able to meet the criteria required for
it. (Quality controls, process followed, and enabling future levels of
test coverage/quality control -- this might have to happen elsewhere.)
Some of the details about "control" of ON are still unresolved, to be
sure, and the current level of Sun control over ON will probably need to
be relaxed somewhat in the future, but at the same time, I expect that
this will have to be met with an increased level of care and process on
the community side, to ensure that the overall quality does not degrade.
>
> The reality is, Sun does not have an infinite number of engineers nor
> an infinite number of resources for testing. Solaris will be developed
> for certain SPARC and x86/x64 platforms, and that code will be released
> into the OpenSolaris tree.
Again, you're making assumptions which may or may not hold true.
>
> Thus once Community-provided platform code is accepted back into the
> OpenSolaris tree, Sun engineers will do what they can to not break
> the Community ports, but Sun will also not do any testing of the
> Community platforms nor any maintenance of them. That will be the
> responsibility of the Community.
Maybe, maybe not. While the ports are not part of ON, this is certainly
true. Which is why New Ports (ports still in incubation, and not part
of ON) makes more sense to me.
>
> So while that does mean a perpetual maintenance train, it's also rather
> similar to the way Linux is maintained. The core Linux maintainers
> maintain the x86 tree, and it's up to the leaders of the various port
> communities to maintain the associated code for ARM, MIPS, PPC, SPARC,
> etc.
I actually abhor the way Linux does this. Its not a good model for
robust engineering IMO. I far, far prefer the NetBSD approach to
portability and architecture support. (And yes, I've real experience
with both approaches.) Anyone who's worked with NetBSD will probably
tell you that their portability and cross-platform support is probably
the one feature that beats all the other FOSS projects, hands down. I'd
far rather aim in that direction, than in the fragmented approach that
Linux has used over the years. But that's just *my* opinion.
>
> That is the model we intend to emulate for OpenSolaris with the
> "Community-Supported Platforms" Community.
Maybe that's the model *you* intend, but I'm not convinced everyone
shares your opinion here. Hmm... were you the original proposer for
the community? If so I guess you have more say in it... but I'd ask
you politely to consider that other models may be superior, or may be
desired by the community. A community straw-poll might be worthwhile here.
-- Garrett
