Jon Mayo wrote:
> It seems to me that everyone has lots of great ideas. It also seems
> to me that a lot of people focus on what I/O pins are not exposed or
> what peripherals are not attached.
Ummm, perhaps because people keep trying cram "mad" designs down its
tiny little throat? :)
> Every exposed pin is going to either increase the cost of the board,
> increase the size of the board or both.
> Every chip added is going to dramatically increase size and cost.
> I am going to assume we are all here for gumstix because of it's
> cost. It's really a good deal compared to other dev kits. And many of
> you are attracted to gumstix because of it's small size.
These assumptions are not strictly true, and who says that by breaking
out a few more signals won't double their sales volume by allowing more
projects to be done without horribly ugly hardware hacks? Wouldn't it
be reasonable to say that the lack of such "extra" board costs might
actually be hurting their sales, because of what is "almost" possible?
In other words, I'd gladly pay $50 more per unit if it meant getting
twice as many people using the platform. So again, I do not agree with
your assertions that such modifications are unreasonable requests, even
with their added costs. These are often called "trade-offs", and the
finding right combination usually takes several tries before finding the
sweet spot. I've just been trying to suggest other possible trade-offs.
After all, the speculation raised about additional I/O directly resulted
from the simple fact that the Gumstix can not readily accomplish what we
might desire today without resorting to gross hacks. Sound, video, and
really high speed buses are not ideal project ideas for the stix,
because you will be duplicating lots of transistors that are already
on-board the PXA and, thus, adding undue and unreasonable costs to your
Moreover, I believe the cost of exposing the I/O off the board will be
*drastically* less than what I imagine will be the case for many of said
hardware hacks in the first place. If you end up spending twice the
time developing a shoe-horned solution, you obviously started with a
flawed foundation for your system architecture.
In other words, Gumstix, Inc. will be saving themselves money at the
expense of their customers development budgets if such projects are
undertaken without reasonable foresight. And I think the continued
consideration of such "mad" solutions completely deserves this
(seemingly pedantic) assessment of the situation.
But hey, this lack of foresight just mean more profits for folks like
myself who offer professional services to bail out folks who find
themselves up such creeks. So go forth and make your messes, and call
me when it doesn't work. My time equals money, and I might stand to
become rich man from such "creative" ideas alone....
> Given that assumption, I don't think we can expect all these great
> I/Os to be exposed. And I'm not going to wait around for BlueTooth
> just so gumstix can have a bunch of the older boards sit around in a
> bin waiting to be sold and end up marking up the BT boards to cover
> the loss.
I agree that there is no reason to wait for new boards, even BT.
Further, I have tried to be expressly clear in my own posts that such
changes will be all but impossible to see realized until well after the
BT model has arrived; however, I again assert that some of the projects
that have been proposed to date will never be economically feasible
until more I/O is available.
While the stix have great potential today, I also pragmatic about the
costs in this business -- and I'm not only talking about the tough
decisions that Gumstix, Inc. itself must consider regarding this topic.
> So I propose that someone, anyone. Put together a list of what we
> know for certain a gumstix can be used for, and place them into two
> categories. The first is if you are a software guy and have only a
> limited ability in electronics. The second if you are a hardware
> hacker and want to slip in every possible peripheral over rs232 and
That seems rather silly. I would rather a single list of projects upon
which hardware and software folks can actually agree. Instead, you
propose a method that will lead to two "extremist views" of what is
possible, without leaving room for an actually tractable middle ground.
In effect, you are proposing the same kind of "unbalanced design" that I
was ranting about in my post about "mad engineering" last week. I'll
pass on that notion, thanks.
Instead, I would rather see the community focus on producing a list of
URL for specifications to chips that *will* seamlessly interface with
the I/O available on the current Gumstik. Without this list, I feel we
are effectively trying to plant a forest without due consideration of
the species of trees that are best suited for the targeted climate.
Once a sufficient library has been built, the Great Ideas should readily
manifest themselves through a strong collaborative effort between the
interested hardware and software folk, and the resulting forest of ideas
will be more immediately approachable. Or so it seems to me.
> Personally I haven't come up with an idea yet, I know it's got to be
> good for something. perhaps with some I2C DACs and things it can
> replace the BasicStamp in my robot. And have some pretty impressive
> mapping functionality. perhaps with some audio-output I can just grab
> a public domain speech synthesizer and have it talk to me (festival
> or rsynth). a rs232 to ethernet adapter and an old DEC or Wyse
> terminal and it can become a cool console workstation. If you add 1
> or two pieces of hardware the possibilities are endless. I hope now
> you understand my comment about "hardware hacker". Now what can we do
> about the software hackers who will be a great help in getting
> kernels, drivers, libraries and applications.
Okay, let me stop you right there. If you were to bring me a board with
an "rs232 to ethernet adapter", I would quadruple the price for the time
I'd spend trying to make it work, as I know enough from past experience
on these matters to know what an utter nightmare you've just proposed.
Sure, today you claim you just want to hook it up to a terminal, but the
second it gets put on a congested network, that design will go up in a
nice virtual cloud of smoke. I personally recommend you steer clear.
That is *exactly* the kind of impractical design idea that makes an
experienced (and sane) software engineer run screaming for the hills --
without ever looking back. Of course, there is always a good mix of
nuts in every professional field, so you can probably find opposing
opinions on this. In fact, I'm sure you can -- I've met said nuts.
> Anyways you all should think about getting a website or project up
> and documented somewhere. Even if you've only just started on it. Put
> some notes on what you've gotten done so far. And once we get a few
> people with projects, either a listing of user websites on
> gumstix.org or gumstix.com would be great. Or possibly just a gumstix
> webring. If anyone wants to coordinate this (or if you want to
> convince me to coordinate it) then speak up now!
I think the webring is a great idea; please consider this my effort to
convince you to start one for the community. I also completely second
the suggestion that everyone begin putting together their own sites, as
such will likely increase the community momentum behind the platform.
This would also allow Gumstix, Inc. to simply reference the ring and
reduce the support burden of updating their list of community links
every time a new site comes on-line. I can't see any downsides.
> ps- From the postings I've seen, I think there is a small (but
> noisy!) few who may have been happier to have a bare PXA on some kind
> of prototyping board (it's probably has too high of a pin count to
> put it on a SurfBoard). Some might have been happy to just have a big
> huge PXA for cheap and hook up all the peripherals themselves. I
> think the rest of us just want a PXA with the ideal combination of
> I/Os to do most, if not everything we want.
First, I really hope you meant to use the word "vocal" rather than
"noisy." The later has some rather negative connotations that I don't
particularly think you intended to convey to individuals like myself.
And if you did, I can gladly demonstrate the difference. :)
Second, there is no "ideal combination of I/O" short of exposing the
pins for every functional module of the processor or giving complete
control (e.g. in an FPGA) over which lines are brought out. Otherwise,
*someone* (me?) will always complain that some other choice should have
been made. Even with all pins exposed, they will still be able to find
reasonable ways of finding faults in the system. Such is life.
There are *always* improvements that can be made to any system, and
those that claim otherwise simply lack sufficient imagination to invent
the next generation of ideas. And I've heard some of the "best" claim
that these types of hardware and software systems *can* be finished,
which did nothing to bolster my opinion of their innovative abilities.
What's really fun is, after you make such drastic changes in response to
demands, the people that asked for them proceed to tell you that they
wanted it changed back -- or aren't willing to pay for them. (I've been
there - fortunately from the software side where it's just a matter
of a rollback - and now require rather hefty deposits before undertaking
projects where such is a clear risk.)
To make it clear that I am not among that wishy-washy crowd, let me
further say that I would put my money down now for a few GumBus cards if
Gumstix, Inc. could give me a date for their arrival. But I do not
expect them to make promises that they may not be able to keep, and any
proposed idea for hardware changes falls into that category since the
boards are already in production.
The only real solution that can ever hope to please everybody is to
offer a family of related products, thus beating the problem that
"nothing can be all things to all people" by having several things that
can be all things to only a good number of people. But again, that is
simply not tractable at this juncture, so we are left with a design that
is well-suited for some things --- and clearly not for others.