Open Source software and OpenSolaris. What is the deal?

(Caveat, much of the following is my OPINION, even when I
don't
explicitly say so. It is also just a starting point for
discussion)

It seems like there is a lot of duplication of effort going
on here.

We have:

sunfreeware.com - /usr/local ??
blastwave.org /opt/csw
opensolaris.org - SFW
opensolaris.org - CCD

It seems the same packages often exist on more than one of
these
repositories, and many times all. (Please let me know if I
missed any
other repositories). Also with the ports community talking
about
setting up their own repository we are getting up to five!

I see opportunity here. Why are we pushing to continue with
a CCD? Why
not throw those efforts into actually making approved SFW
packages,
that will become part of OpenSolaris and eventually
Solaris?

I would like to see the CCD and SFW communities merge. One
idea for
CCD packages, is to make the new spot to introduce packages
into SFW.
These packages would follow the same path, and other
requirements as
SFW. They would also be protoypes for including a new
package into
Solaris. Clearly since these have the potential to impact
namespace,
they would be required to have a full ARC proposal and
review.

Once they have been introduced as a "CCD" package,
they could
relatively easily be considered for inclusion into Solaris
SFW.

Where does this leave sunfreeware.org and Blastwave?
Sunfreeware.org
and Blastwave should combine their efforts. How that would
happen,
requires further investigation.

Maybe the combined effort would result in a /opt/sfw tree
that uses
pkg-get!! (Both would become part of the OpenSolaris effort,
as part
of the SFW community)

In the ideal world this effort would go into the SFW
consolidation,
but let's take it one step at a time. Maybe we can't figure
it all out
in one pass, but we should definitely get the conversation
rolling.

Also the ports collection should consider throwing their
efforts into
the blastfreeware collective.

Cheers,
-Brian

P.S. - I am not really sure what the gnu-sol-discuss group
is for, but
I included it because it seemed relevant.
_______________________________________________
gnu-sol-discuss mailing list
gnu-sol-discussopensolaris.org

Brian,

> It seems like there is a lot of duplication of effort
going on here.
>
> We have:
>
> sunfreeware.com - /usr/local ??
> blastwave.org /opt/csw
> opensolaris.org - SFW
> opensolaris.org - CCD
>
> It seems the same packages often exist on more than one
of these
> repositories, and many times all. (Please let me know
if I missed any
> other repositories). Also with the ports community
talking about
> setting up their own repository we are getting up to
five!

There are a number of reasons for this duplication - some of
it is
historical, some of it is that these "communities"
serve different
interest and some is just organizational.  For example, the
SFW effort
comprises software which is currently supported by Sun
Microsystems in
its Solaris product while the CCD effort incorporates
packages which
are not supported by Sun but are packaged to make it easier
for users
to get commonly-used open source.

> I see opportunity here. Why are we pushing to continue
with a CCD? Why
> not throw those efforts into actually making approved
SFW packages,
> that will become part of OpenSolaris and eventually
Solaris?
>
> I would like to see the CCD and SFW communities merge.
One idea for
> CCD packages, is to make the new spot to introduce
packages into SFW.
> These packages would follow the same path, and other
requirements as
> SFW. They would also be protoypes for including a new
package into
> Solaris. Clearly since these have the potential to
impact namespace,
> they would be required to have a full ARC proposal and
review.
>
> Once they have been introduced as a "CCD"
package, they could
> relatively easily be considered for inclusion into
Solaris SFW.

With respect to CCD and SFW, I've been thinking along the
same lines.
In general, I completely agree with this although longer
term I don't
see a need to introduce any more CCD packages.  Instead what
I would
propose is that all such externally-derived open source be
integrated
into the SFW effort.  CCD could continue to exist for
Solaris 10 but
for future releases, externally-derived open source could
come through
SFW.

> Maybe the combined effort would result in a /opt/sfw
tree that uses
> pkg-get!! (Both would become part of the OpenSolaris
effort, as part
> of the SFW community)

Perhaps pkg-get or perhaps something else but I believe
providing a
repository of SFW "packages" that can be
downloaded via an apt-get-like
facility would be tremendously useful.

dsc
_______________________________________________
gnu-sol-discuss mailing list
gnu-sol-discussopensolaris.org

I have a slightly different take on this.

BTW: I have a little head start here because Brian and I had
a fairly
lengthy e-mail exchange on this.  I'm really glad that more
are in on the
discussion now.

David.ComaySun.COM wrote:
> Brian,
>
>> It seems like there is a lot of duplication of
effort going on here.
>>
>> We have:
>>
>> sunfreeware.com - /usr/local ??
>> blastwave.org /opt/csw
>> opensolaris.org - SFW
>> opensolaris.org - CCD
>>
>> It seems the same packages often exist on more than
one of these
>> repositories, and many times all. (Please let me
know if I missed any
>> other repositories). Also with the ports community
talking about
>> setting up their own repository we are getting up
to five!
>
> There are a number of reasons for this duplication -
some of it is
> historical, some of it is that these
"communities" serve different
> interest and some is just organizational.  For example,
the SFW effort
> comprises software which is currently supported by Sun
Microsystems in
> its Solaris product while the CCD effort incorporates
packages which
> are not supported by Sun but are packaged to make it
easier for users
> to get commonly-used open source.
>
>> I see opportunity here. Why are we pushing to
continue with a CCD? Why
>> not throw those efforts into actually making
approved SFW packages,
>> that will become part of OpenSolaris and eventually
Solaris?
>>
>> I would like to see the CCD and SFW communities
merge. One idea for
>> CCD packages, is to make the new spot to introduce
packages into SFW.
>> These packages would follow the same path, and
other requirements as
>> SFW. They would also be protoypes for including a
new package into
>> Solaris. Clearly since these have the potential to
impact namespace,
>> they would be required to have a full ARC proposal
and review.
>>
>> Once they have been introduced as a "CCD"
package, they could
>> relatively easily be considered for inclusion into
Solaris SFW.
>
> With respect to CCD and SFW, I've been thinking along
the same lines.
> In general, I completely agree with this although
longer term I don't
> see a need to introduce any more CCD packages.  Instead
what I would
> propose is that all such externally-derived open source
be integrated
> into the SFW effort.  CCD could continue to exist for
Solaris 10 but
> for future releases, externally-derived open source
could come through
> SFW.
I don't think the distinction of "supported" and
"not-supported" goes away
that easily.

For example, there are probably a lot of folk who use KDE on
Solaris and
even more that would like too.  Given more "Window
System" engineers,
I think we would use them to better support Gnome than
additionally support
KDE. Some things just won't be supported (by Sun, who knows
what
might happen under OpenSolaris, but they are hard pressed to
provide
support).

I think the concept of SFW is right on (although the
implementation is a
bit messed up and often misunderstood).

I think the concept of CCD is simply bad.  We should not be
providing
a recompilation and packaging service. We should be
providing assistance
to the ultimate code maintainers to provide Solaris
"packages" just like 
they
provide RPMs (or whatever) for Linux. Let's turn our
"recompile" junkies
into "Solaris Packaging Evangelists" (easier said
than done - different
skill set - but hopefully you get the idea).

In addition to the "Sun must limit its support
burden", there are a couple
of reasons for this.

First, it puts us on fairly level ground wrt the more
conservative Linux
distros - the ones that lock down on a version for the life
of the 
distribution,
delta fixing really serious problems.  (The obvious analogy
is "like Red 
Hat,
not like Fedora".)  You get stability between well
defined "flag 
points", plus
you have the option of putting newer (perhaps not supported)
versions on
the system (directly from the maintainers).

Second, there is an unavoidable latency introduced by
placing a "CCD"
service in the loop.  The Linux distros don't have this
latency.  Having
such a latency will probably kill us. Part of the
"charm" of FOSS isn't
that its particularly stable, but the rapidity with which
fixes to affected
components appear when instability happens.  For
non-Sun-supported
software, we need to cash in on that rapid turnaround.
>> Maybe the combined effort would result in a
/opt/sfw tree that uses
>> pkg-get!! (Both would become part of the
OpenSolaris effort, as part
>> of the SFW community)
>
> Perhaps pkg-get or perhaps something else but I believe
providing a
> repository of SFW "packages" that can be
downloaded via an apt-get-like
> facility would be tremendously useful.
I couldn't agree more that an improved repository, hosted
(hopefully) by
OpenSolaris (ignoring who ultimately pays the bill), would
be a great
thing.  I only differ (I think) on where those packages come
from. I
think we should host packages built by the maintainers.

Its probably worth it to play with Ubuntu or Debian here. 
They clearly
decompose the packages they offer along the lines of license
and support.
After installing Ubuntu, the first thing I do (because I'm
old and my eyes
aren't what they used to be) is update the nVidia driver. 
They make it
clear (but painless) that I'm downloading something from
them, they aren't
going to directly support.  Its also clear that its not true
FOSS for the
religiously inclined. (Note: It was always possible, but a
bit more
painful in earlier Ubuntu releases.  I suspect its gotten
less painful based
on customer feedback. (Hummm, maybe I should ask my Ubuntu
buddy
about that.))
> dsc
- jek3

_______________________________________________
gnu-sol-discuss mailing list
gnu-sol-discussopensolaris.org

Joseph Kowalski wrote:
> I think the concept of CCD is simply bad.  We should
not be providing
> a recompilation and packaging service. We should be
providing assistance
> to the ultimate code maintainers to provide Solaris
"packages" just like 
> they
> provide RPMs (or whatever) for Linux.

Wearing my hat as a code maintainer for one of the major
open source
upstream communities (X.Org), we provide the same set of
packages for
all OS'es - those are source tarballs only.    We don't
provide RPM's,
debs, gentoo ebuilds, SVR4 packages, or any sort of binaries
- there
are simply too many different versions of different distros,
with
different package systems, dependencies, supported
platforms, to even
begin.   The only thing we (or many other upstreams) can
support is
source release, and expecting each distro to provide it's
own packages.
We don't have a warehouse to fill with the hundreds of
machines necessary,
nor do we have the number of volunteers necessary, to
provide packages
for any OS.

I just don't see trying to push package building upstream as
a viable
solution at all.

-- 
	-Alan Coopersmith-           alan.coopersmithsun.com
	 Sun Microsystems, Inc. - X Window System Engineering
_______________________________________________
gnu-sol-discuss mailing list
gnu-sol-discussopensolaris.org

On 5/4/07, Peter Tribble <peter.tribblegmail.com> wrote:
> On 5/4/07, David.Comaysun.com <David.Comaysun.com> wrote:
> > With respect to CCD and SFW, I've been thinking
along the same lines.
> > In general, I completely agree with this although
longer term I don't
> > see a need to introduce any more CCD packages. 
Instead what I would
> > propose is that all such externally-derived open
source be integrated
> > into the SFW effort.  CCD could continue to exist
for Solaris 10 but
> > for future releases, externally-derived open
source could come through
> > SFW.
>
> I don't like this. The problem with it is that if it
follows existing practices

Keep in mind by having Blastwave's S11 consolidation,
following
OpenSolaris' it will be possible for an Sun Solaris 11
customer to
type: sfw-get upgrade-latest gawk, and have it write over
the included
version. (We get rid of patches, and just upgrade packages)
Some other
option might be upgrade-base, and upgrade-unstable

> Now, either we can get to a state where we can rapidly
integrate new
> versions of software into the stable SFW tree very soon
after their
> release (which seems unlikely, given that much of the
software we're
> talking about doesn't have strong stability
guarantees), or we need
> a separate track into which the new versions can be
injected. And yes,
> this will often mean that there are multiple versions
of the same app
> on the system at the same time.

Software is vetted through Blastwave S11.

> So I think we need at least 2 different efforts. What
the second
> one ought to be based on I'm not sure.
>
> (And yes, even for the software I myself build and
maintain,
> I usually find myself maintaining multiple versions -
one for
> stability and maintaining dependencies, and a second to
get
> all the new features, and maybe others.)

Exactly... I suspect the SFW and coolwave maintainers will
eventually
have a good deal of overlap.

-Brian
_______________________________________________
gnu-sol-discuss mailing list
gnu-sol-discussopensolaris.org

> Again, lets hold off on making these decisions until
the draft
> document is ready. I have already accounted for these
issues.
>
> What I am most concerned about at this point is the
delivery of
> software rather than those that maintain it. Not that
this isnt
> important, but it seems we are putting the cart before
the horse so to
> speak.

Huh? Software delivery is not really what we are talking
about here. This is:

http:
//docs.google.com/Doc?id=df6hcjb2_13hrtqsz

Note that Software delivery is bullet 6 out of 9. We have
other things
to work out. You should start a new thread with a different
name to
discuss software packaging and delivery. (Vs. mixing it up
with this
thread.)
_______________________________________________
gnu-sol-discuss mailing list
gnu-sol-discussopensolaris.org

On 04/05/07, Brian Gupta <brian.guptagmail.com> wrote:
> On 5/4/07, Peter Tribble <peter.tribblegmail.com> wrote:
> > On 5/4/07, David.Comaysun.com <David.Comaysun.com> wrote:
> > > With respect to CCD and SFW, I've been
thinking along the same lines.
> > > In general, I completely agree with this
although longer term I don't
> > > see a need to introduce any more CCD
packages.  Instead what I would
> > > propose is that all such externally-derived
open source be integrated
> > > into the SFW effort.  CCD could continue to
exist for Solaris 10 but
> > > for future releases, externally-derived open
source could come through
> > > SFW.
> >
> > I don't like this. The problem with it is that if
it follows existing practices
>
> Keep in mind by having Blastwave's S11 consolidation,
following
> OpenSolaris' it will be possible for an Sun Solaris 11
customer to
> type: sfw-get upgrade-latest gawk, and have it write
over the included
> version. (We get rid of patches, and just upgrade
packages) Some other
> option might be upgrade-base, and upgrade-unstable

I don't see how.

Bastwave is what it is because it has its own isolated
software stack.

It is not built such that the dependences included with
Solaris 10 are
used in the build process (as far as I know).

It is not designed to replace the base system.

I'm also very unsettled by your proposal that this system
should
arbitrarily try to undo Sun official patches, packages,
etc.

There are many of us that do NOT want a package system that
is
designed to replace the base system's packages.

-- 
"Less is only more where more is no good." --Frank
Lloyd Wright

Shawn Walker, Software and Systems Analyst
binarycrusadergmail.com - http://binarycrus
ader.blogspot.com/
_______________________________________________
gnu-sol-discuss mailing list
gnu-sol-discussopensolaris.org

Peter,

>> With respect to CCD and SFW, I've been thinking
along the same lines.
>> In general, I completely agree with this although
longer term I don't
>> see a need to introduce any more CCD packages. 
Instead what I would
>> propose is that all such externally-derived open
source be integrated
>> into the SFW effort.  CCD could continue to exist
for Solaris 10 but
>> for future releases, externally-derived open source
could come through
>> SFW.
>
> I don't like this. The problem with it is that if it
follows existing 
> practices
> with SFW, then the stability constraints rapidly result
in fossilization.

First of all, the SFW I'm talking about is similar but not
the same to
the SFW we've had in the past.

To clarify, I don't believe the effort being run out of SFW
precludes
differing stability constraints.  In fact, my assumption is
that there
will be software of various stabilities integrated through
the
consolidation.  Certain pieces of software - for example,
libxml - may
change more slowly than other pieces of software and will
not change
incompatibly.  Other pieces of software might change
incompatibly at a
Minor release boundary (for example, SunOS 5.11) but then
would not
over the lifetime of that particular Minor release.  And
some pieces of
software may change much more rapidly.

> Now, either we can get to a state where we can rapidly
integrate new
> versions of software into the stable SFW tree very soon
after their
> release (which seems unlikely, given that much of the
software we're
> talking about doesn't have strong stability
guarantees), or we need
> a separate track into which the new versions can be
injected. And yes,
> this will often mean that there are multiple versions
of the same app
> on the system at the same time.

I don't believe SFW precludes supporting this sort of
effort.  Previous
efforts have tended to deliver software which was fossilized
but there
is recognition that this needs to be addressed.

> So I think we need at least 2 different efforts. What
the second
> one ought to be based on I'm not sure.
>
> (And yes, even for the software I myself build and
maintain,
> I usually find myself maintaining multiple versions -
one for
> stability and maintaining dependencies, and a second to
get
> all the new features, and maybe others.)

A number of us have been discussing as a requirement the
ability to
support not only multiple versions of software but also
multiple
streams of change - this would allow certain users to ride
the
slow-moving train which changes slowly over time and large
and/or
incompatible changes come at very well defined points.  But
it should
also allow a class of users who desire much more rapid train
- for
example, a user accustomed to the Solaris Express biweekly
release.
Even more risk tolerant users might choose to live on the
"nightly"
train.

dsc
_______________________________________________
gnu-sol-discuss mailing list
gnu-sol-discussopensolaris.org

Alan Coopersmith wrote:
> Joseph Kowalski wrote:
>> I think the concept of CCD is simply bad.  We
should not be providing
>> a recompilation and packaging service. We should be
providing assistance
>> to the ultimate code maintainers to provide Solaris
"packages" just 
>> like they
>> provide RPMs (or whatever) for Linux.
>
> Wearing my hat as a code maintainer for one of the
major open source
> upstream communities (X.Org), we provide the same set
of packages for
> all OS'es - those are source tarballs only.    We don't
provide RPM's,
> debs, gentoo ebuilds, SVR4 packages, or any sort of
binaries - there
> are simply too many different versions of different
distros, with
> different package systems, dependencies, supported
platforms, to even
> begin.   The only thing we (or many other upstreams)
can support is
> source release, and expecting each distro to provide
it's own packages.
> We don't have a warehouse to fill with the hundreds of
machines 
> necessary,
> nor do we have the number of volunteers necessary, to
provide packages
> for any OS.
>
> I just don't see trying to push package building
upstream as a viable
> solution at all.
>
Uh, isn't X.org the exception rather than the rule?

If I have this backwards (and most communities only provide
source), then
my apologies to all.

I have noted that few provide "dpkg" format.  You
need to go to Debian or
Ubuntu to get that.  Could it be that RPM just occupies a
special, 
historical
nitche?

- jek3

_______________________________________________
gnu-sol-discuss mailing list
gnu-sol-discussopensolaris.org

> It seems like there is a lot of duplication of effort
going on here.

IMHO, Sun (the entity that runs OpenSolaris from behind the
scenes) should be spearheading the effort to consolidate and
make mature these critical areas of OpenSolaris.  If there
is one thing they want to get right the first time, it is
package management.  And I'm not sure the community is
capable of solving this problem without turning things into
a hodgepodge of competing but undifferentiated standards
(like Linux).  

Ubuntu is the most successful open source desktop operating
system of all time.  It has strong leadership pushing
critical decisions down from the top, thus minimizing the
classic wasted efforts found in Linux history, while still
capitalizing on open software development.  If I was Sun,
I'd look there for direction and inspiration.  

That said, they've probably done exactly that by hiring Ian
Murdock to lead the Solaris strategy.  So maybe he'd have
something interesting to say about this topic.
 
 
This message posted from opensolaris.org
_______________________________________________
gnu-sol-discuss mailing list
gnu-sol-discussopensolaris.org