Topics

my 5 cents on AGL and Tizen


Alison Chaiken <alison@...>
 

Automotive is important to the future of Linux, and AGL is important
to the future of automotive Linux, so please take my comments in that
spirit. I *personally* support the AGL initiative and wish it
success. Having said that, I share many of the concerns expressed by
Todd Fuchs, Jeremiah Foster and Pete Popov.

Moblin was a decent, pretty distro. MeeGo was even better in many
ways, as evidenced by its continuing support by many vendors. Both
of those projects were, in the end, failures. Tizen-IVI is Intel's
new automotive initiative, and AGL is largely Tizen. Let us learn
from the mistakes of the past so as not to repeat it!

I don't have a problem with Tizen or Android as reference distros.
They both are featureful, well-supported, under rapid development and
modern in design. I notice that no one in AGL mentions Android as a
possibility. Why not? I'd guess because Android is different in
important ways from plain vanilla Linux, even as mainline unifies the
Android kernel with the upstream one, as the userspaces are quite
distinct. Also, Android is under the control of one vendor, and its
governance body is not open. Now compare Tizen: looks very similar!
And recall that Moblin and MeeGo failed in part because they were
too much under the control of a few companies whom outsiders did not
quite trust.

I have used OBS, and I am impressed by EFL. However, these are not
standard user-space choices any more than a Java-based runtime is.
The idea that developers just want to code in HTML5 is short-sighted,
as the many startups I talk to here in Silicon Valley want mostly to
create telematics applications that offer services like remote door
opening and state-of-charge reporting. I can't believe having
followed the LayerManager and AudioManager and RadioTuner discussions
in GENIVI that coders who want to talk to hardware (which is *most* of
them) will be able to achieve the functionality they want using just
HTML5. HTML5 would need significant extensions to work with GENIVI
components.

Contrary to what others have asserted, there is tremendous startup
interest here in automotive: notice that the Silicon Valley Automotive
Open Source Group I founded now has over 500 members, many of whom are
entrepreneurs, angel investors or large adjacent businesses looking to
make an acquisition. The hottest field is usage-based insurance,
which is highly dependent on data from vehicular buses and ECUs.

Why are Ubuntu and Android so popular? Because they are
ISO-installable on standard hardware and have well-documented and
useful build tools and maintained repos. Everyone knows that in
most cases, Ubuntu or Android will "just work" rapidly, with minimum
cost. What AGL should offer in order to be wildly successful is an
installable ISO that runs under QEMU and AWS and has reasonably priced
reference hardware with both x86 and ARM architectures. The ISO
would include build tools, source packages and demo apps.

If AGL is just a rebadged Tizen-IVI whose important discussions take
place on tizen.org mailing lists, it will follow Moblin and MeeGo into
the dustbin. I don't make this statement because I wish it to be so,
but because I'm certain of it. At a bare minimum, AGL *must*
support a plain vanilla Linux variant as CGL did, one that is not
based on OBS or EFL. The Yocto-IVI layer produced by Holger
Behrens and his team at Wind River strikes me as one obvious choice.
Note that WR is a direct competitor to my employer, so my endorsement
is based purely on a technical evaluation, not self-interest.
Assuredly Mentor's or MontaVista's or Wind's plain-vanilla Linuxes
would all be fine, as would Debian.

The competition to Linux in automotive is QNX and Windows. Both of
these are complete, simple-install bundles. AGL must be as easy to
use in order to compete.

I will not be at CES, but will attend Embedded Linux Conference and
Linux Collaboration Summit, so let's talk if you will be at one of
these conferences.

Happy New Year,
Alison

--
Alison Chaiken alison@she-devel.com
650-279-5600 http://{she-devel.com,
exerciseforthereader.org}
A car is just four wheels, a few sofas, plus a lid and an engine. --
Li Shufu, founder of Geely Motors
via Michael Dunne's _American Wheels, Chinese Roads_


Juha-Matti Liukkonen <juha-matti.liukkonen@...>
 

Hi!

Alison, you make very good points. Let me comment on a couple specific ones.

On 1.1.2013, at 1:27, Alison Chaiken <alison@...> wrote:

 What AGL should offer in order to be wildly successful is an
installable ISO that runs under QEMU and AWS and has reasonably priced
reference hardware with both x86 and ARM architectures.    The ISO
would include build tools, source packages and demo apps.

Something to this effect is what we have been thinking as well. Also documentation plays an important role in making adoption easy: the adaptation layer (for HW) and HMI APIs (for apps) need to be well documented, with examples to get started from. And of course the run-time layer in between needs to be robust.

 At a bare minimum, AGL *must*
support a plain vanilla Linux variant as CGL did, one that is not
based on OBS or EFL.     The Yocto-IVI layer produced by Holger
Behrens and his team at Wind River strikes me as one obvious choice.

I do not disagree with these statements. Yocto-IVI was very close to being the initial base for AGL work. There were reasons why Tizen-IVI was chosen as the initial reference implementation; time will tell how that works out.

It is also important to remember that the AGL initiative output is not distro-specific. The goal is to improve open source components that make up the platform, such that they are better suited for automotive use, going upstream first. Tizen may be our reference test and integration platform, but the results can certainly be integrated to e.g. Yocto based builds.

Best regards,
- Jussi

--
Juha-Matti Liukkonen
Technology Director, Embedded software


Rudolf Streif <rstreif@...>
 

Thank you, Alison, for your contribution. Discussions like these help to shape AGL and it is good that we have them. We are just starting out with AGL and it will take many minds and hands to make it successful.

My vision of AGL is summarized in these bullets:
  • to enable the automotive industry to successfully leverage Linux and open source in product development;
  • to close technology gaps through contribution to existing open source project or to launch open source project where existing ones cannot be leveraged;
  • to provide a reference distribution that allows engineers to ramp up quickly so that they do not have to jump through endless buildsystem and dependency hoops to make it work;
  • to provide reference hardware on which aforementioned reference distribution has been tested;
  • to provide good documentation;
  • to provide education so that engineers from the industry will be able to eventually contribute and successfully work within open source projects;
  • to provide education on open source licensing, compliance, supply chain management and how to work with your own IP in the context of open source.
I do agree with you that AGL's reference distribution needs to be an installable ISO that is well tested and runs out of the box on the reference hardware and hardware very similar to it and on QEMU. It needs to be very easy to use and very easy to adopt and port to new systems. I want AGL to be a full Linux OS stack that comes with great documentation and tooling so that can be as easily adopted and ported to new systems. It is the ease of use that makes Android so successful on mobile systems. People are talking a lot about Android in automotive systems but they are concerned about its technical limitations. I see AGL to be for automotive what Android is for mobile but without these limitations.

AGL needs a starting point. It does not make sense to spend a lot of resources to do yet another integration effort for a distribution. Tizen, building a system with the Yocto Project, WebOS, WRT, etc. are all valid options. However, none of them is where what AGL ultimately needs. They all have their unique advantages and disadvantages. Any Linux distribution is a collection of packages that are built and compiled into an image using a choice of tooling.

The distribution is a starting point because it will only get developers that far because sooner or later they are confronted with the task of modifying the distribution and adapting it and adding their own packages. For that they need the tooling and education on how to use it.

Reaktor and Jussi Liukkonen stepped up and provided instructions on how to get started with Tizen IVI and using the build tools. Pete Popov and Todd Fuchs (and probably others) picked it up and are now experiencing hurdles getting from source code to an image. I hope that we can improve the instructions based on those experiences. That is what we have the wiki for.

I would also love to see someone step up and provide instructions on how to build a stack using the Yocto Project and provide them on the wiki. The end result would be very similar to Tizen IVI since most Linux stacks are very comparable and it would provide developers with the opportunity to compare the different approaches of building a Linux software stack. I do not see it as conflict or too much fragmentation to have different ways of building AGL software stacks. In terms of package selections they can converge. They will all serve as a platform to build and improve on but cater to different needs of building and integrating.

In the end all of these stack will have the same technology gaps that AGL needs to close to make them ready for automotive prime time. All of these stacks will serve as a great platform to improve on. However, our focus must be on improving, documentation, tooling, etc. and not that much whether one solution is better than the other. There is no one-size-fits-all.

We do need people to step up and do actual work. We can discuss all day long but it is action and commitment that make the difference. In that spirit, kudos to Jussi, Pete and Todd who are actually doing that.

Rudi



On Mon, Dec 31, 2012 at 3:27 PM, Alison Chaiken <alison@...> wrote:
Automotive is important to the future of Linux, and AGL is important
to the future of automotive Linux, so please take my comments in that
spirit.     I *personally* support the AGL initiative and wish it
success.  Having said that, I share many of the concerns expressed by
Todd Fuchs, Jeremiah Foster and Pete Popov.

Moblin was a decent, pretty distro.    MeeGo was even better in many
ways, as evidenced by its continuing support by many vendors.   Both
of those projects were, in the end, failures.    Tizen-IVI is Intel's
new automotive initiative, and AGL is largely Tizen.     Let us learn
from the mistakes of the past so as not to repeat it!

I don't have a problem with Tizen or Android as reference distros.
They both are featureful, well-supported, under rapid development and
modern in design.    I notice that no one in AGL mentions Android as a
possibility.    Why not?   I'd guess because Android is different in
important ways from plain vanilla Linux, even as mainline unifies the
Android kernel with the upstream one, as the userspaces are quite
distinct.    Also, Android is under the control of one vendor, and its
governance body is not open.    Now compare Tizen: looks very similar!
   And recall that Moblin and MeeGo failed in part because they were
too much under the control of a few companies whom outsiders did not
quite trust.

I have used OBS, and I am impressed by EFL.   However, these are not
standard user-space choices any more than a Java-based runtime is.
The idea that developers just want to code in HTML5 is short-sighted,
as the many startups I talk to here in Silicon Valley want mostly to
create telematics applications that offer services like remote door
opening and state-of-charge reporting.    I can't believe having
followed the LayerManager and AudioManager and RadioTuner discussions
in GENIVI that coders who want to talk to hardware (which is *most* of
them) will be able to achieve the functionality they want using just
HTML5.   HTML5 would need significant extensions to work with GENIVI
components.

Contrary to what others have asserted, there is tremendous startup
interest here in automotive: notice that the Silicon Valley Automotive
Open Source Group I founded now has over 500 members, many of whom are
entrepreneurs, angel investors or large adjacent businesses looking to
make an acquisition.   The hottest field is usage-based insurance,
which is highly dependent on data from vehicular buses and ECUs.

Why are Ubuntu and Android so popular?    Because they are
ISO-installable on standard hardware and have well-documented and
useful build tools and maintained repos.    Everyone knows that in
most cases, Ubuntu or Android will "just work" rapidly, with minimum
cost.    What AGL should offer in order to be wildly successful is an
installable ISO that runs under QEMU and AWS and has reasonably priced
reference hardware with both x86 and ARM architectures.    The ISO
would include build tools, source packages and demo apps.

If AGL is just a rebadged Tizen-IVI whose important discussions take
place on tizen.org mailing lists, it will follow Moblin and MeeGo into
the dustbin.   I don't make this statement because I wish it to be so,
but because I'm certain of it.     At a bare minimum, AGL *must*
support a plain vanilla Linux variant as CGL did, one that is not
based on OBS or EFL.     The Yocto-IVI layer produced by Holger
Behrens and his team at Wind River strikes me as one obvious choice.
 Note that WR is a direct competitor to my employer, so my endorsement
is based purely on a technical evaluation, not self-interest.
Assuredly Mentor's or MontaVista's or Wind's plain-vanilla Linuxes
would all be fine, as would Debian.

The competition to Linux in automotive is QNX and Windows.    Both of
these are complete, simple-install bundles.     AGL must be as easy to
use in order to compete.

I will not be at CES, but will attend Embedded Linux Conference and
Linux Collaboration Summit, so let's talk if you will be at one of
these conferences.

Happy New Year,
Alison

--
Alison Chaiken                           alison@...
650-279-5600                            http://{she-devel.com,
exerciseforthereader.org}
A car is just four wheels, a few sofas, plus a lid and an engine. --
Li Shufu, founder of Geely Motors
via Michael Dunne's _American Wheels, Chinese Roads_
_______________________________________________
automotive-discussions mailing list
automotive-discussions@...
https://lists.linuxfoundation.org/mailman/listinfo/automotive-discussions



--
-- 
Rudolf J. Streif
Director of Embedded Solutions
The Linux Foundation

Phone: +1.619.631.5383
Skype: rudolfstreif
PGP: RSA 2048/2048 D6E7D28B

Linux Foundation Events Schedule:  events.linuxfoundation.org
Linux Foundation Training Schedule: training.linuxfoundation.org