Danek Duvall wrote:
> I think we've gotten to this statement by a circuitous, and not altogether
> sound, chain of reasoning. The actual problem, as I see it, is that the
> ARC has started to get inundated with a whole bunch of cases that don't fit
> the rules that we've been working by for the last twenty years, and the
> output has generally been "Syntax Error".
+1
Part of the problem is trying to define the problem correctly, and Danek has
(as usual) done a good job of articulating it in a way that doesn't conflate
it with IPS. Thank you.
Several of us talked about this quite a bit today (in person) and came to the
same conclusion that Danek and Stephen did. The only role IPS plays in this
conversation is as a catalyst - the "multiple repositories" mantra (which is
similar to the "multiple consolidations" one) is a convenient red herring.
That doesn't mean that there isn't a grain of truth in it - in this case,
truth is that we need a way of identifying which things need to get that
extra burst of effort needed to integrate them tightly into OpenSolaris,
which things will only get a little TLC, and which things won't get any
effort at all. Thinking of it as "different repos are the magic bullet",
however, was lazy and misleading thinking. Sorry.
Where are we now? IMO:
The Goal:
We need to maintain the (Open)Solaris value propositions of interface
stability over time, robustness and enterprise-quality integration,
while at the same time innovating and including as much of the FOSS
world as we can.
The real problem statements?
Doing a good job of integrating FOSS projects into OpenSolaris can
be a costly and non-trivial effort,
Reviewing thousands of projects that don't wish to spend any effort
to integrate into OpenSolaris is a waste of time and effort, both for
the ARC and the project teams,
Most FOSS projects were written to run on Linux. In as much as
OpenSolaris does some things in a different way than Linux, simply
compiling these apps on Solaris doesn't guarantee that they will
work, much less be well integrated and use the "native OpenSolaris"
mechanisms,
As we add more and more FOSS, the differentiation between "native
OpenSolaris" and "Linux" interfaces blurs and goes away.
If we mix poorly integrated components into our OpenSolaris source
trees in an undifferentiated manner, we will gradually lose the
ability to construct a system that only includes tightly integrated
components, and
If we don't find a way to provide easy access to the majority of FOSS
components in OpenSolaris, developers and users will slowly stop
using OpenSolaris.
One of these new processes might well be "tagging projects with an
expected integration level" so that we can apply the right amount of
resources and review to each project. Whether that tag is an ARC-only
fiction, or whether it is an IPS repo tag is completely irrelevant.
Another tool might be JohnF's FOSS Checklist, but it, in itself, is
much too heavyweight to scale to 10,000 projects/year, simply because
we don't have the resources (in the Community or Sun) to fill out that
many checklists.
There may be more.
"This project is Purple. It gets the Purple kind of review. This other
one is Green, and so it gets handled like all those other Green projects.
It is OK to treat Purple projects different from Green ones."
-John
