RPi use cases explained

Here is my take on this odd situation.  There are several needs and types of stakeholders in the Pi ecosystem:

• Conceptual need in EE.  Eben Upton has many times related the problems he experienced while doing student recruitment  at university, where each year's intake candidates were seemingly  less technically experienced than the last.  I can well believe that, because we saw exactly that same problem in my own engineering department, to the point where we had to provide catch-up courses to bring part of the student intake up to the level where they could  understand basic 1st-year EE lectures.  Making very cheap hardware available that encourages experimentation by inquisitive youngsters does seem to address part of the problem squarely by fostering interest and offering direct experience with hardware.  However, it doesn't address the gap in mathematics and foundational science skills.

• Improving IT education.  The UK has a specific problem in school-level IT education, in that over time it became nothing more than vocational training in office skills.  Clearly there is much room for improvement there, but this is almost entirely unrelated to the skills shortage observed in EE recruitment.  CompSci departments might benefit a little if programming were taught in schools, but not a lot because lack of programming skills is not the bottleneck,  Stronger maths skills would be vastly more useful than programming knowledge, and would help EE as much as every other branch of engineering and the physical sciences.  Also, programming is almost always vocational training with just a smidgeon of CompSci education acquired by osmosis on the side, and very rapidly becomes dated.  To compound matters further, a high-level language with a lot of abstraction would tend to be chosen for programming education, which means that pupils would tend to learn little about computer fundamentals unless they have an awesome teacher who explains the foundations along with the programming.

• Cheap media centre.  Don't laugh, this is a major stakeholder group for Pi.  What's more, RPF have always known this, because they have promoted the very strong media capability of the Broadcom SoC countless times in their blog.  They even went as far as to sell licensed codecs which are about as distant from educational as anything could be.  This area may well be getting the most development effort as well, which is reasonable since it plays to the Pi's biggest strength and makes a very large group of Pi users happy.

• Platform for expansions.  It always did seem odd that the Foundation so often stressed the difficulty of reaching their $25/$35 price point, and yet created a board bearing proprietary MIPI DSI and CSI-2 connectors which raised the board cost and complicated PCB routing.  Even more odd is that these MIPI interfaces would not contribute significantly to the board's educational capabilities since USB cameras and displays with open interfaces were readily available at good prices.  The subsequent high investment by RPF in developing camera and display modules suggests that this was a planned business strategy from the start, and it explains why the extra connector cost was considered justified.  One possible view is that there is business advantage in creating a platform for which expansion modules could be produced using a proprietary interface spec that narrows the competition.  Whether or not that was the thinking, it is the current actuality since RPF has invested time and money in expansions and delivered product.

• Enthusiasts/makers hacking platform.  Quite distinct from the needs of EE and UK IT education, a  large group of stakeholders is the worldwide and ever-growing community of makers and related enthusiasts, which may or may not be technical.  This group is heavily interested in creative projects which typically underpin some other area of interest that isn't itself computing.  The Foundation has from the start shown some interest in supporting this group, as evidenced by the board's P1 interfacing header and the near-miraculous provision of SoC peripheral interfacing information from a SoC manufacturer that has shown very little interest in supplying open documentation.  The enthusiast/maker community is strongly aligned with the open source software and open hardware communities since closed/proprietary devices impede rather than support building things.  Unfortunately RPF has been lukewarm in this area as the board is not open hardware, the SoC has very little open documentation, and not all of the software is open source either.  Undoubtedly most of the blame for this lies with Broadcom, but RPF spokespersons have defended the restriction of information themselves as well.

• Commercial for-profit product.  This stakeholder group is small but obvious.  RPF is a registered non-profit, but Premier Farnell and RS are not, and so the Pi has to justify its place on warehouse shelves.  The typically high profits on accessories probably make this quite easy though.

It's pretty clear from the above that the Pi ecosystem has multiple interested parties and drivers, and proceeds along many roads simultaneously with varying degrees of support from the Foundation.

In other words, the Raspberry Pi's concept, rationale, targets and user base are not correlated.

Getting a single view from anyone (especially RPF) is no more productive than all those blind men feeling different parts of the elephant.  To say that it was designed for IT education is completely wrong if intended literally --- Pi would not have been designed as it was nor targetted so strongly at non-educational stakeholders if that had been the primary intention.  This makes the question of "Why is there still no educational release?" a simple one to answer:  IT education was only one driving force, and clearly not a major one.

The only certain position is from objective engineering:  it's an ARM board with specific pros and cons, and it's those pros and cons that determine its effectiveness or otherwise for any given application.  I guess that's more boring than hype-laden official positioning statements by people with vested interests, but in contrast to them, it's accurate.

Morgaine.

Morgaine Dinova wrote:

 

• The subsequent high investment by RPF in developing camera and display modules suggests that this was a planned business strategy from the start, and it explains why the extra connector cost was considered justified.  One possible view is that there is business advantage in creating a platform for which expansion modules could be produced using a proprietary interface spec that narrows the competition. 

Actually the proprietary nature of the interface doesn't seem to be much of a problem. The actual camera sensors with the same interface and physical plug appear to be reasonable easily available off ebay and such like as spares for phones.

The roadblock is the bits that are buried inside the GPU.

I'd have to say that I originally thought the camera was a daft idea, but it's cheap enough that I bought one anyway.  If JamesH manages to sort out a couple of software niggles, I can see it having a reasonable future as a very cheap megapixel security camera amongst other things.

I see someone has already produced an aluminium case with a mount for the camera and externally a mount for additional standard lenses.

I think that's where the maker community shines.. by taking these cheap component parts and combining them into all sorts of interesting ideas.

Morgaine Dinova wrote:

 

• The subsequent high investment by RPF in developing camera and display modules suggests that this was a planned business strategy from the start, and it explains why the extra connector cost was considered justified.  One possible view is that there is business advantage in creating a platform for which expansion modules could be produced using a proprietary interface spec that narrows the competition. 

Actually the proprietary nature of the interface doesn't seem to be much of a problem. The actual camera sensors with the same interface and physical plug appear to be reasonable easily available off ebay and such like as spares for phones.

The roadblock is the bits that are buried inside the GPU.

I'd have to say that I originally thought the camera was a daft idea, but it's cheap enough that I bought one anyway.  If JamesH manages to sort out a couple of software niggles, I can see it having a reasonable future as a very cheap megapixel security camera amongst other things.

I see someone has already produced an aluminium case with a mount for the camera and externally a mount for additional standard lenses.

I think that's where the maker community shines.. by taking these cheap component parts and combining them into all sorts of interesting ideas.

The camera interface is in fact a standard not unlike the display the problem here is that you can talk to the device but you don't know what to say to it when you have got it. Each camera has it's own set up which is really where it becomes a true pain.

Other than that the closed source GPU is quite the pain more so than the camera really since it controlls so many other subsystems

John Alexander wrote:

 

Each camera has it's own set up which is really where it becomes a true pain.

 

Other than that the closed source GPU is quite the pain more so than the camera really since it controlls so many other subsystems

The problem here, of course, being that the driver for the camera is in the GPU. So even if you have all the info on the camera there's nothing you can do with it.

I've been following the section of the RPF forums on the camera and some people seem to have obtained most, if not all, of the relevant info on the camera and who are coming up with questions detailed enough to have stumped the broadcom employees working on the camera...

Combined with people who appear to have a lot more photographic knowledge than the software people it's made for some interesting reading.

At best it played to the SoC's strengths by providing a camera interface with more bandwidth than the Pi's heavily shared USB can provide, but it offered nothing fundamental that can't be done in other ways, and adding those connectors didn't help keep the cost down nor contribute towards the skills level of undergraduate applicants.

It's consistent with everything else about Pi, lots of directions and stakeholders.  I don't think that people looking for a clear statement of requirements followed by a corresponding clear effort in that direction (or even directions) will find it.  Pi didn't happen that way, and after nearly a year and a half of watching RPF's efforts, it's easy to see that their main investment hasn't been towards the much discussed and over-hyped educational goals.

I think it was coder27 who said that their main concern seems to have been to sell a lot of boards.  Well we know that they have, but ignoring the benefit of hindsight, it's also clear that they have put a large amount of effort into media support and expansion modules, so "selling product" must indeed rank near the top of their own list of goals.

Along a rather different tack ...

If the main aim were charitable support of education, the price of Pi would have dropped by now anyway.  Their BOM cost now is just a fraction of what it was in the days of 10k-30k volume costing, so if they wished to lower the barrier further, I bet they could, and very easily.   (Farnell and RS are likely to be making nice profits on accessories anyway.)

Morgaine Dinova wrote:

the price of Pi would have dropped by now anyway.  Their BOM cost now is just a fraction of what it was in the days of 10k-30k volume costing, so if they wished to lower the barrier further, I bet they could, and very easily.

I suspect that bringing manufacturing back to the UK hass offset that somewhat, but I agree.  If they'd somehow limited it to education and excluded the makers and xbmc folks they wouldn't have had the sort of volume pricing though...

I suspect that nobody would argue much if they simply said that the price stays the same for everyone else and that they'd use that to give education a nice discount, it'd at least go some way to showing a bit more conviction on the educational side.

If the main aim were charitable support of education, the price of Pi would have dropped by now anyway.

I think the answer to that is pretty clear.  They like the price point and want to use Moore's law to increase

the capabilities rather than further lower the price.

Pete Lomas is quoted as saying recently in a 14 May 2013 interview:

Maintaining Pi's progress is important. "What we don't want is for Pi to be a one trick pony," Lomas said. "The educational goal is too important. So our focus continues to be on the Pi's educational value. How can we improve that? Would the addition of more computing power increase the educational value? Would the introduction of more accessories make a difference?"

...

Nevertheless, Lomas admitted the Foundation is thinking about Pi2 – what it might be and how it might fit with Pi1. "Moore's Law says there will be a better processor and there will be more memory and that will allow us to maintain the price."

http://www.newelectronics.co.uk/electronics-technology/raspberry-pi-foundation-looks-to-the-future/50316/

I thought that

"The chip has a powerful gpu and we want that to be used for educational purposes – there's no point in having a gpu if you can't use it."

Lomas' view is that vision will be a key learning experience. "People need to learn how facial recognition works, for example," he said. "The module will bring a new level of opportunity."

was an interesting comment..

The Arm just isn't powerful enough to make a lot of this practical and squeezing the raw images through the gpu->arm interface seems to be a bottleneck as people trying to side-step the GPU's image processing are finding out.

The GPU not being open enough means you only get to use what the firmware provides. I'm not sure of the educational value in simply being able to tell the GPU to do the dirty work for you if you can't get inside that process and play with how it works yourself.

Pete Lomas is quoted as saying recently in a 14 May 2013 interview:

 

"The educational goal is too important. So our focus continues to be on the Pi's educational value.

A line seems to have got lost:

"So we will continue to focus on everything except the educational release."

Would the addition of more computing power increase the educational value?  -- Lomas

Lomas' view is that vision will be a key learning experience. "People need to learn how facial recognition works, for example," he said.

I think Pete Lomas would make a great salesperson for BBB, with its increased

computing power and support for machine vision.  Until Pi2 (Raspberry Tau?)

comes out, there seems to be no good reason to push RPi for education.

"So we will continue to focus on everything except the educational release."

That's consistent with the question yesterday:

I am seeking input about enterprises that are using Raspberry Pi for internal applications to meet internal needs.

http://www.element14.com/community/thread/25484

There is a linked-in page for someone with the same name which says:

I write invaluable case studies and white papers that accomplish technology companies’ marketing and messaging goals.