USB update from Gordon, new direction

That sounds like a far more promising direction, as getting low-latency realtime operation from a 1-core CPU running a Linux kernel was always destined for disappointment, a design fault in fact.  The VPU in contrast seems to have oodles of power.

Unfortunately, pushing the USB work into the VPU will make Pi a more closed platform than it is currently.

Morgaine Dinova wrote:

 

...Unfortunately, pushing the USB work into the VPU will make Pi a more closed platform than it is currently.

A "horses for courses" compromise, I think. Most Pi customers would probably settle for closed low level stuff that_just_worked so that they could concentrate on the higher level applications that they were already trying to implement. Those with the chops to do low level hacking are probably looking to better documented / more proven hardware already (unless they're very stubborn...)

It sounds like a bit of a **** or bust move to me (asterisks mine btw). In any case it's an interesting move away from the well-trodden "pesky Linux timing" get out. Time will tell whether tossing data at a fast-in-a-parallel-kind-of-way-but-not-really-expecting-this-sort-of-job section of the silicon will bear fruit.

Jonathan Garrish wrote:

 

It sounds like a bit of a **** or bust move to me (asterisks mine btw). In any case it's an interesting move away from the well-trodden "pesky Linux timing" get out.

Well they don't really have much in the way of options.  They're currently releasing a partially broken product, and given their preoccupation with image at all costs, they probably want to fix the USB by hook or by crook as a priority.  Well the hook won't work because Linux is not a realtime kernel, so the crook of doing USB handling in the VPU is their only real alternative.

Replacing the BCM2835 by a different SoC with a fully operational USB is also an alternative in principle, but I suspect not a realistic one at this stage.

How do they expect the community to help with a piece of software that runs on an undocumented piece of hardware?

I also asked the question on their Forum how much time they expect this new development strategy to take.

I never got an answer on that. Creating an usb host implementation from scratch seems like a terrible lot of work to me, specially if it's done by a couple

of people in their spare time.

In the mean while, people believe again the issue will get solved and sales keep going.. 

Luc Cool wrote:

 

How do they expect the community to help with a piece of software that runs on an undocumented piece of hardware?

I also asked the question on their Forum how much time they expect this new development strategy to take.

I never got an answer on that. Creating an usb host implementation from scratch seems like a terrible lot of work to me, specially if it's done by a couple

of people in their spare time.

In the mean while, people believe again the issue will get solved and sales keep going.. 

Regardless of the availability of documentation it's still a big ask - all the information in the world won't help if there's maybe only a dozen people in the known universe that could thow an implementation together in a timely manner. Even more so if those dozen people don't have to rely on a successful implementation in order to eat.

I'm seeing apologists over at the other place referring to the Pi as an ongoing beta and inferring that it's caveat emptor for anyone who buys a Pi for any purpose other than as a low cost command line educational tool and I feel it puts the whole open source philosophy in a difficult position:

1) The Pi is piggybacking on the Linux community, but only because it's a convenient and free (as in beer) operating system. Linux might be getting wider exposure as a result, but I'm not sure that exposure will be positive in the long run. If I were one of the Keepers Of The Kernel I'd be telling the Foundation to pull their fingers out, lest I publically distance my product from their sillyness. Unfortunately everybody is free to use Linux, even in really unflattering implementations...

2) Many Linux users welcome the Pi as a cheap bit of hardware and are happy to ignore / make excuses for lack of documention / the proprietary technology / some woeful software and hardware problems. This selfishness doesn't help customers who find that their new piece of hardware isn't as good as everybody says it is and then get short shrift when they ask for solutions.

I'm no open source evangelist, but I do recognise the importance of the movement as a foil against the excesses of "the other guys". The trouble (and it's manifesting itself with the whole Pi thing) is that very few within the movement appear willing to criticise anything vaguely connected with open source - even if it's patently problematic. It's easy to rail against [enter name of multinational here], or to stroke one's beard while waffling about the finer points of GNU, but to the 99% of people who just want hardware / software that doesn't suck then it's arcane and irrelevant. A bit more pragmatism required perhaps

And no, we're not all in the position to "fix it ourselves" - even if it is all open and documented (and let's face it - a lot of open stuff suffers from lack of documentation / commenting and the whole github etc model is out of the ark). Even the smartest of us will at some point have to call upon the services of a plumber, mechanic, lawyer, or other specialist because we want the job done right. Waiting for someone in the community to come fix that burst pipe in the roof - for free and in their spare time - will only end in disappointment (and mouldy carpets).

Someone must be making some money from 1/2 million sales of the Pi. Maybe it should be up to them to fix the issues (or at least publicly state what it can or cannot do). But that will never happen so long as the apologists have their way.

USB will work when they find the formula to create milk out of a rock. I believe I mentioned it long time ago, but the piece of IP that Broacdom got from Synopsys for the BCM2835 is an implementation of USB OTG, it will be very challenging to pretend make it work as a full featured USB HOST.

Can you do some workaround to get some limited functionality working properly for some devices, probably yes, but it will never reach Eben's promise that you can do with the Pi what you can do with any other computer, that has been a misleading statement since day way as it is the permanent tag message on the Blog header.

Still the Rpi is a cool gadget to thinker with Linux.

-J

jamodio wrote:

 

...Still the Rpi is a cool gadget to thinker with Linux.

 

-J

I would venture to say that ~most~ people don't have a computer in order just to "tinker" with it's operating system - although some of us will doubtless have spent far too much time poking about in the innards because something doesn't work, or just out of curiosity...

Nope, for most a computer is a box for doing stuff - interfacing with a sensor, browsing, coding, playing media, writing that novel and all the other wonderful / mundane things. "Tinkering is fine, so long as it's not compulsory!" It shouldn't be compulsory either - not everyone wants to be a dentist, or a lion tamer - even though they are doubtless noble professions.

Ironically, for those folks that are keen on destruction testing an operating system then the best choice of hardware IMO is something absolutely bulletproof. Then, when the rainbow / blue screen / spinning beachball of abject misery appears one know's it's directly related to one's own input and not some random hardware / driver burp that one has no control over. That's why that with hindsight the original choice of Fedora for the Pi seems a bit ...umm... foolhardy. Beta hardware + alpha software = debugging purgatory.

Noticed that I said "gadget" not "computer" ...

Jonathan Garrish wrote:

Beta hardware + alpha software = debugging purgatory.

Amen

jamodio wrote:

 

Noticed that I said "gadget" not "computer" ...

 

Noted and understood!

Actually it's worse than that, more like:

Beta hardware + alpha software + closed source + restricted doc = not a chance