[yocto-ab] Introducing IoT Reference OS Kit

Philip Balister philip at balister.org
Fri Feb 24 11:49:13 PST 2017


On 01/23/2017 04:10 PM, Cobbley, David A wrote:
> Hi Philip,
> 
> Thanks for the thoughtful questions.  I’ll answer inline below.
> 
> --David C
> 
> From: Philip Balister <philip at balister.org>
> Date: Friday, January 20, 2017 at 7:53 AM
> To: David Cobbley <david.a.cobbley at intel.com>, "yocto-ab at yoctoproject.org" <yocto-ab at yoctoproject.org>
> Subject: Re: [yocto-ab] Introducing IoT Reference OS Kit
> 
> David, this is great news.
> 
> As an advisory board member, I'd like to understand what this mean the
> Yocto Project is committing to though. Are these specific layers for
> building existing software, new software packages, etc.
> [DAC} This is a set of layers maintained in their own repo, just as Poky is published now. Some recipes would obviously be referenced/reused from other layers, while this would contain some new recipes for (x) along with their dependencies (y), the point being that we would want to stick to the proven format for repository and release that already works for Poky
> 
> 
> We already have one reference distribution (Poky). How does this differ?
> [DAC] Poky is designed to be as general purpose as an embedded (reference) distribution can possibly be, to be used as a learning tool and as a starting point. If someone were to base a product on Poky, and many people do, they need to modify it to their own requirements and then add their own application on top. IRK by contrast is a much narrower reference kit designed for a specific purpose, so much of the grunt work is done for specific IOT devices/gateways.
> [DAC]  We also see the potential for other reference kits in the future that would contain different types of recipes. IRK, for example, contains support for industrial robotics. Maybe an Automotive Reference Kit (in cooperation with AGL?) might contain CAN bus support. Other vertical ref kits would presumably contain support for standards or features relevant to those specific verticals.
> 
> Finally, we need to understand what this does to the manpower needed to
> support the Project. Richard is constantly reminding us we need more
> people working on core pieces of the project, will this lead to
> additional work for already overloaded people?
> [DAC] Excellent concern. Intel is donating both the software and the maintainership of this reference kit, so no additional work would be required by the project. As stated in another response, we are taking on the full burden of the kit as defined today, not asking the community to support this.  Those who want to extend/expand this kit for other domains or purposes, or create their own reference kits, would also own the incremental work.
> 

I've been thinking about the general reference kit problem.

I'm very uncomfortable with having a Yocto Project reference kit driven
by only one member, especially when I see that the concept should
benefit most, if not all members. So I'd really like to here from other
members how they feel about creating a useful format for reference kits.
Clearly, part of this motivation is to expand the base of companies
maintaining the kits.

Since my expertise is SDR, I'll describe the SDR reference kit.

The heart of this kit is the meta-sdr layer, so the required layers are
oe-core, meta-sdr, and several other layers with dependencies (that I
forget).

The Yocto project test matrix for the layer would be machines defined in
oe-core so we can guarantee the kit builds.

The individual companies interested in using the kit would be
responsible for run time testing on their hardware (or testing with
their specific distribution configuration).

So the reference kit concept (from the Yocto Project point of view) is
provide a set of build tested layers that serve as base for Company
specific solutions.

Before the Yocto Project accepts a reference kit, I'd like to see the AB
agree what the purpose of a reference kit is and how they are structured
for easy re-use by all members.

I realize I'm late replying to this thread, but we need to get this right.

Philip

> 
> Philip
> 
> On 01/19/2017 04:27 PM, Cobbley, David A wrote:
> As Yocto Project focuses over time to better support new and distinct verticals (to attract a greater membership), we wanted to help that effort by offering a reference distribution that enables key components and technologies relevant for IoT. The IoT Reference OS Kit is a new set of Yocto Project metadata layers and infrastructure geared towards IoT usages. The reference kit will introduce the concept of a "profile" - an image configuration that integrates a subset of the new content allowing the content to work seamlessly together. Examples of planned profiles are industrial, gateway and machine vision, but others could be included. Each profile implements certain validated key use cases.  The reference kit would follow the Yocto Project release cadence, publish the content in the Yocto Project git and use the Yocto Project Bugzilla for feature and bug tracking.
> Initially, the code which will be contributed by Intel will support selected Intel development platforms through the meta-intel BSP layer, but other contributors are welcome to port and test the reference kit on additional architectures, or perhaps create distributions specific to other verticals. Intel will maintain the IoT Reference OS Kit content relevant to Intel Architecture, and Yocto Project will benefit from the project through expansion of the amount of well-maintained content. The reference kit will provide an additional validation mechanism for the large part of OE-Core as well as extending the testing to metadata in other layers and we would naturally contribute relevant improvements and bug fixes we come across with during our development cycle also to core Yocto and OE-Core and those other layers as appropriate.
> Some planned high level features for the reference kit are:
> -OTA SW update mechanism
> -3rd party application support
> -Comprehensive sensor support (mraa/upm/IoTivity)
> -BT Audio
> -Industrial robotics support
> -Machine Vision support including Realsense enabling
> -Security enhancements like TPM,IMA/EVM and Secure Boot reference implementation
> We are excited to contribute these new IOT-friendly features to the Yocto Project, and will be releasing the initial implementation into the project soon.
> --David Cobbley
> Intel SSG/Open Source Technology Center
> 


More information about the yocto-ab mailing list