Open Hardware Architecture

[bbrv]

Hi, we thought it might settle some of the tension by posting this idea here.  We have also posted it on other sites to generate discussion.

Very soon Genesi plans to introduce the following:

__OpenPowerArchitecture__ ...or OPA

An Open Hardware/Software System Architecture Reference

1)   Mission

a. The mission of OPA is to provide a flexible and scaleable hardware abstraction layer for PowerPC driven devices. The scope is to cover a wide range of devices, ranging from Smart Phones to SMP server blades. As a well defined interface between hardware and software OPA will bring benefits to both sides of the system design. Building OPA compliant hardware will instantly enable this system to run all the OPA compliant software (e.g. various operating systems available for the OPA, including LINUX, BSD, QNX, VxWorks and others). The hardware vendor has no longer the need to develop or license various support packages for all the different operating systems.  Likewise, software developers do not need to attend to all the requirements of all possible platforms. Supporting one OPA system means supporting every OPA system. This means that even a specialized design of custom hardware that is built in a total quantity of ten units has instant access to all the operating systems and software that are OPA compliant. Using OPA will bring down system development time and cost in reasonable way while speeding up time to market. OS vendors will be enabled to support systems at release time by providing one OPA version.

b. Learning from the mistakes of CHRP, PREP and POP...etc.
 
c. Strategic concept: Economies do not grow strong by restricting the flow of information; they grow strong by sharing ideas and building upon the work of others. In fact, all of mankind’s greatest inventors built upon the work of their predecessors. There is no work that has not built upon the previous knowledge of others at all; we could not write these words were we unable to build upon the lingual knowledge that has preceded our existence upon this globe.

2)  Concept of the Method -- Defining the OPA interface

This will be used as the only software/hardware interface of a OPA compliant system. A OPA system:

 HARDWARE<->FIRMWARE<->OPA interface<->OS

Please note that in the meaning of OPA the firmware is treated as hardware part of the system. All pieces used to build the OPA interface are built out from open standards and free of any patents and licenses. Any IP created outside of known and freely available standards (like PCI, IEEE1275, etc.) should by trademarked, patented or otherwise secured in the name of OpenPowerArchitecture.org. or OpenPowerArchitecture.com (we have registered both) then can license this IP with at symbolic fee of $1 ( -- with "copyleft" agreement). We have actually thought about placing this in a Foundation (and have had a suggestion of an endowment!).

OPA will have a total of three function sets:

a. Minimum required functions:

The minimum functions for a OPA system include a device tree, IO access, system control, etc. These are to be used by the OPA OS.

b.  Optional functions

These OPA functions are hardware implementation and/or system vendor specific. As an example, an OPA hardware implementation of a embedded system like a smart phone or a game boy type of device has no need of SMP control functions as needed for a multi CPU server blade. These are to be used by the OPA OS.

c. User functions

While maintaining the calling conventions of OPA functions, this type of functionality are fully hardware vendor specific and can be used to control special custom hardware, etc. The OPA OS will never use these functions, but will provide a API to allow special applications accessing these functions.

3) Business Model

a. OPA.org provides the standard itself and does the needed licensing stuff at various levels (e.g. single user, commercial customer, schools, non profit organizations, etc).

b. OPA.com provides reference implementations of the OPA standard that can be used for evaluation, design and verification. These implementations are available for licensing to enable a customer to build an OPA compliant system without be required to do a complete development from the beginning.

This includes:

i)    a hardware reference design

ii)    a firmware reference design

iii)    a OS reference design

Any feedback?

R&B  :-)

[bloodline]

iii)    a OS reference design

AROS

[Cyberus]

bloodline wrote:

iii)    a OS reference design

AROS

Right on cue.... :-)

[hnl_dk]

ii)    a firmware reference design

U-Boot would be perfect :-D It would be great with an opensource Firmware for an open orchitecture :-)

[Karlos]

Not really a signifigant point, but isn't OPA the 'familiar' German word for Grandpa?

I may have something rather more intelligent to contribute after I've got over this raging headache...

[DanDude]

a gamebox, yes.

Amiga, NO.  

You're putting software developers out of business if you do this.

[Cymric]

If I understand correctly you want to create a standardised abstraction layer ('OPA') for any PowerPC-based computer platform, the rationale being that such hardware offers a consistent interface to operating systems. The latter can then pick and choose what functions of the interface it will use or build upon. And, since all compliant OSs just see one type of hardware (namely OPA), coding would become more efficient, faster, and so forth. Barring in mind that I have no real experience with CHRP and do not even own a PowerPC-based system... My first gut reaction was 'why??!'

Do you expect such a diversity of PowerPC-based systems that such a standard is necessary?
Do you expect that all OSs you mention will be made OPA-compliant, and if so, in what time frame and by whom?
How do you plan on defining the standard? (I'm thinking W3C here, which according to many vendors is so slow that companies implement their own extensions and more or less get them standardised after the entire world is using them 'illegally'!)
Is there a solid reason why an OS designed for, say, a mobile phone, could or should be run on, say, a fully-equipped PC running Linux, or vice versa? (Think Windows[0-9]+ vs. WindowsCE.)
Is there even sufficient common ground between all types of hardware to warrant such a single interface?

Basically, I'm trying to figure out the answer to the age-old question: if you can do something, should you do it?

[bbrv]

Hi Cymric, good questions!

While Linux has its own hardware abstraction layer, it always gets joined to the machine's BIOS or Boot Prom through a loader, be it LiLo, SiLo, MiLo or Grub. One model we have been envisioning is similar to the SiLo model, which is the way the OS speaks to Sun hardware. The bootstrap process takes place through commands given the boot prom, then control is handed off back to the kernel when that process is complete.

The overall point being made is that for once, all programmers will have access to the internals of the firmware and the physical machine without having to be registered with a company and paying constant fees to get that knowledge and the updates that come with it.  This would be geared to the Open Source market.  We need them and we get a hardware sale.

Microsoft and Intel have been talking about a modified kind of BIOS that (supposedly) takes care of Security concerns, but in reality, is a MS tactic to keep other OS's from being installed on those machines. In retaliation, the Free software community has accelerated efforts to make a BIOS of their own that can be integrated into new hardware no matter what the Wintel coalition does.  We need to offer this first.  The Pegasos/O.P.A. thus becomes an even more important weapon in that "struggle," and it will draw a lot of attention and attract we think many talented allies.

Was thinking about the "Open" community and different layers of use...:-)

BTW, MorphOS is something completely different.  OPA is a way to promote the Pegasos as a reference design.

R&B  :-)

[reflect]

c. Strategic concept: Economies do not grow strong by restricting the flow of information; they grow strong by sharing ideas and building upon the work of others. In fact, all of mankind’s greatest inventors built upon the work of their predecessors. There is no work that has not built upon the previous knowledge of others at all

If you look at what has been accomplished in the world of science, you also know that the only breakthroughs that has been made, was made cause some individuals ignored old knowledge and overlooked "facts"..  they thought outside of the box. Leonardi Da Vinci is an excellent example of this, having no education, taught himself most of what he knew and reasoned in ways not hampered by traditional knowledge/facts.

[bbrv]

True, but Latin was still useful as well as a few few other notable examples!  :-)  He started out apprenticed to the renowned workshop of Andrea del Verrochio in Florence, as a painter (working inside a 2D box so to speak).  He ended his life as the "Premier Painter and Engineer and Architect" of the King, Francis I of France.  He did alot of other stuff too!

BTW, the first Swedish Pegasos User Meeting is happening right now in Gothenburg!  You should be there!  ;-)

R&B  :-)