[Arm-netbook] Regarding Improv, EOMA68, Free Software and Open Hardware.

[Henrik Nordström]

tor 2014-06-26 klockan 21:47 +0100 skrev Luke Kenneth Casson Leighton:

>  and you know what? i'm actually relieved, because this is supposed to
> be a decade-long standard.  if we'd had the funding, we would have
> gone ahead, the EOMA68 standard would have SATA and it would have
> about one SoC available per two years *or less*.

Agreed.

>  the changes i've made since it started open it up to 23or more SoCs
> *per year*.  i am also considering - reluctantly - reserving the 5.0mm
> card height for 1280x800 capable SoCs and the 3.3mm card height for
> 1920x1080 SoCs (over the 24-pin RGB/TTL that is)

Why? I don't get it.

For 1920x1080+ LVDS / DSI / eDP is much better suited interfaces imho.
Unfortunately all (including RGB/TLL, but except eDP) is plauged by
configuration issues and vendor extension dependencies which makes then
somewhat troublesome for EOMA.

> > As for Improv, by the time it entered (the hardware) it provided nothing
> > unique that developers could not find elsewhere. Also most people in my
> > humble opinion already had given up on EOMA68 as a serious initiative
> > due to the numerous problems that project had already seen. I still back
> > the basic ideas of EOMA68, but not sure about the current realization of
> > it, aiming purely at low end market in interface specifications.
> 
>  but henrik, this is entirely entirely missing the point.  the entire
> point for end-users is that you buy 1 CPU Card and share them across
> products, saving 30 to 40% for doing so *and* having consistent user
> applications and data across all products.

No I am not missing the point at all.

My point is that Improv is a development platform. EOMA68 is at this
point in time a niche product. The combination of the two makes an
extremely small market for Improv. The primary group that Improv targets
have other choices that fulfill their needs better.

EOMA68 needs something like Improv.

The majority of the users Improv targets do not need EOMA68 or even
benefit from it.

> > Also,
> > it has to be realized that any such specification do have a fairly
> > limited lifespan and needs to be revised regularly, which somewhat
> > nullifies the benefits.
> 
>  no.  the interfaces were picked because they have all been around for
> over a decade, and they are anticipated to be around and in
> mass-production use for at least another decade.

Agreed that the selected interfaces is going away any time soon, but
these specifications forces the lower end of the spectrum, which also
makes it an uninteresting path for the tech-sawy people you need to get
the whole thing flying.

>  do you see USB being retired in the next 10 years? (serious question)

No. But RGB/TTL is likely fading quickly as display density increases.

>  with over 2,000 active LCD panels on panellook.com do you see RGB/TTL
> being retired in the next 10 years?

Outside the low-end product spectrum yes, obviously.

>  likewise I2C, likewise GPIO, likewise SD/MMC, likewise SPI, likewise
> UART, likewise ethernet.

No, but I question our ability to today define which of these is really
needed.

>  ... except doesn't take a stand on GPL violations, and distributes
> illegal copyright-violating product, but _apart_ from that they
> "believe in OSHW", yes.

It's a hardware company, not a software company. Their software
development resources are very limited, and do take these infringements
seriously.

>  no i am fully aware that the CPU Cards have "less interfaces" quotes.
> that is entirely deliberate, as the interfaces had to be:

And I was not critizing EOMA there.

>  in this way we can ensure that even the lowliest SoC (for example the
> IC1T which is barely struggling to meet the absolute minimum of EOMA68
> requirements) stands a chance.

Is that at all a desireable goal?

Timewarp two years, would such device have any reuse value at all by
then?

Same applies to the other half of EOMA, the base-board (or perhaps
better named base-device). The EOMA idea only makes sense if pieces are
up to a level that reuse and upgrade makes sense.

Or in other words, any EOMA device designed today need to be designed
with the goal of being valuable to reuse in at least 2-3 years time.

Any device manufactured at the lower range of todays standard is very
unlikely to have any meaningful value in two to three years in the same
consumer group.

>  if we had put e.g. CSI or TS on, or eDP or MIPI, then only the
> absolute absolute top-end SoCs could be included,

today yes, unless you accept to throw in a converted chip, which is hard
doe to the extreme tight space requirements in the EOMA standards.

>  look at the example of SATA.  i wanted it - everyone wanted it - but
> in the end it had to go, because it's not lowest-common-denominator
> (and USB3-to-SATA ICs can do a good enough replacement job).

Heck, even storage over USB2 works well for most practical uses today in
the target focus of EOMA.

>  you need to tell the improv team that, not me.  remember, they were
> just a 3rd party client, so we were waiting for the cash order.  all
> the money is held by that 3rd party.  so you need to tell *them*, not
> me, ok?

And I did. As stated in the first paragraph this message was a public
copy of my response to their letter with only minimal edits to fit the
mailinglist.

> > I'd
> > rather see you focus on moving forward than how to cover the refunds.
> > Failures are all natural path of moving forward and learning.
> 
>  true.   ain't quitting.

Neither am I. Just temporarily held up in other completely crazy
business and learning a lot more about hardware side of things than I
ever thought I would be doing.. and mixed together with free software.
But almost entirely outside the scope of this mailinglist even if we did
cause noticeable shortage of Beaglebone Blacks on the market..

Regards
Henrik