Openmoko build / R&D vs Integration

[Richard Purdie]

On Wed, 2008-04-16 at 22:24 +0100, Andy Green wrote:
> There are two subthreads in one here, one on each of those topics, so it
> can seem confused.
> 
> 1) "development": my suffering trying to get a handle on building DM2
> with a library dependency with bitbake confirming  my downer on OE, and

You have this "downer" on OE yet you don't seem to want to try and
understand the several options presented to you to address this and fix
the problem?

> 2) "distribution": would abandoning OE for Moblin / Ubuntu Mobile, or
> going in another direction still, leave us in a better place after the
> dust settled.

Its an interesting question but its very hard to evaluate on the
information available.

> The question that has turned up in front of us courtesy Intel is: do we
> want to trade off some of that "different, hard to learn" against the
> wide spectrum of used and unused options that "powerful" gives us?

Ok, what exactly is this offer? You're asking people to contrast OE
against something that doesn't yet exist and has unknown configuration.

You are surrounded by people who know something about building
distributions for embedded systems and their opinion seems pretty clear
yet you don't seem interested in them. They know the problems OE has
overcome and the features it has, this new system is a total unknown.

Whats needed here is a detail explanation of what exactly this Intel
offer is, how long it will take, how it would work and what the
responsibilities of the various parties are. You can't even say if it
would compile natively or cross compile, just that you'd trust Intel to
get it right.

> I spoke several times about Openmoko not being able to afford to be in
> the R&D game.  The only game we can play successfully is "integration",
> to take chipsets and software and drivers from external vendors or
> projects and figure out how to get them to work in one product.  See the
> Glamo as what happens when we play R&D with even drivers.
> 
> (When I first joined I went as far with this logic as to ask why are we
> making our own GUI apps at all.  The answer was it's our special sauce,
> but now I see Qtopia turning up.)

Every time there is one of these direction changes there is a say 6
month (or worse) hit to what the community sees getting developed.
Qtopia isn't the first and by the looks of things, the lesson hasn't
been learnt and people are still trying to pull things in many
directions at once. Focus is important.

> Today Openmoko essentially does distro R&D: but we could take up Intel's
> offer and merely integrate distro generation and maintainence from
> "upstream" and plug into Moblin / Ubuntu Mobile with a connected sugar
> daddy helping us along.  Naturally we would work closely with that
> upstream and maintain and provide packages, etc as made sense.

I don't see what R&D you do now for distro maintenance that you won't
have to do with Intel's offer. I don't think you'll get something on a
golden plate that has everything out the box, I see a lot of work in
tweaking whatever is provided to make it suite OpenMoko's needs and that
will probably be heavier than your current workload IMO.

A good question is why would Intel be offering this? The cynic in me
says they're unlikely to be doing it out of the goodness of their heart.
Their motive will govern where the project goes and is it in the same
direction as Openmoko? With OE as a build system you can pretty much
choose yourself since OE has a policy of no such constraints, other
systems may not have such policies.

Also, experience shows that whatever system you develop, they do take
time to develop and stabilise regardless of how many developers you
throw at it. Maemo makes an interesting case study of a system with
success. I'd measure the time it took in years. Also note there is no
way to build a maemo system, just the apps, one at a time. For every
success you have lots of vapourware...

Anyhow, I've said more than enough. I've watched the Openmoko project,
I've admired it for what its trying to do, I've tried to help it on its
way where I could (bitbake SRCVREV for example) and if Openmoko feels
another change in direction is what it needs, so be it. You would
probably argue everyone is closed minded and not prepared to consider
the Intel proposal seriously, equally you come across as closed minded
towards OE which you appear unprepared to even try to use.

Regards,

Richard