Kuro, MediaMVP, TiVo, TeraStation -- and us?

Georg Wrede georg.wrede at nospam.org
Fri Nov 3 15:37:15 PST 2006



Gregor Richards wrote:
> Georg Wrede wrote:
> 
>> While on the IBM Developer Works site, I stumbled on an interesting 
>> article about the new breed of computing devices (IMHO, soon to flood 
>> the world, in numbers exceeding both PCs and cellphones).
>>
>> http://www-128.ibm.com/developerworks/linux/library/l-lobmediamvp.html
>>
>> Basically, these are VCR cassette size units containing a single-board 
>> computer with Linux. They seem to be (again, IMHO) becoming the new 
>> platform-of-choice for garage firms developing totally new 
>> applications for the digital world. (Gee, sounds like I'm selling them!)
>>
>> Many new firms use these as-is in home heating automation, various 
>> set-top appliances, for automotive control, security systems, 
>> robotics, toys, point-of-sale terminals, self service trade show 
>> stands, firewalls, handheld bookkeeping or measuring devices, 
>> computers for pre-schoolers, etc.
>>
>> As for myself, for the better part of this year I've been looking for 
>> something smaller than a real PC, because I need to deploy several 
>> computers throughout the factory floor, and for this PCs get clunky 
>> and waste a lot of room, money, and maintenance resources. Some of the 
>> choices I've eyed at are:
>>
>> http://www.armkits.com/Product/sbc2410III.asp
>> http://www.armkits.com/Product/STDV710A.asp
>> http://www.harerod.de/centipad/html/centipad.html (in German)
>>
>> Now, what has held me back so far is, I definitely want to continue 
>> all my development in D! <putting hand behind ear, waiting for intense 
>> applause and roar from the audience/>
>>
>> With the GCC toolchain that comes default with most of these, one 
>> would assume that GDC is a viable alternative.
>>
>> Opinions?
> 
> 
> GDC is certainly a viable alternative. With ARM, you're already set. 
> configure-make-install and you're done. Cross-compilation is easy.

Excellent!

> Seems like a non-problem to me.

What sort of executable sizes can I expect from GDC for ARM? (Last time 
I tried, DMD Hello World was monstrous compared to C on Windows and 
Linux.) With limited space this may become an issue when one has to have 
a bunch of subtask-specific executables. (Which one would have when 
doing stuff "in the spirit of Unix". I wouldn't want to resort to 
Busy-Box like binary crunching methods with virtually separate executables.)

> PS: We all know that VMs with Java are the ideal solution for small, 
> low-memory low-power boards. Because VMs rool.

Of course. Then I could go off on a tanget about

  - using only one language for _all_ development on all platforms in 
this project

  - ease of development without bothering with JVMs and Java libraries, 
and possibly other "best-for-this" languages for each task and platform

  - using one language only for everything eases maintenance, 
documentation, in-project portability, human cost, reliability, 
transparency, reviews, debugging, accountability, ....

  - D being a Systems Language and therefore demanding it's good enough 
for this

But frankly, I _want_ to use D, and I recognize this fact. (As opposed 
to hordes of people I've known in various places, who also _want_ 
something, but are unable to recognize this for themselves, and instead 
rationalize (often very convincingly) that choice to everybody, 
including themselves.)

And of course, while D does have a size problem right now, it really 
shines in output per clock tick, which also is important here. The size 
problem I trust will go away with the forever plummeting media costs. 
But for me, today, in this project, it still is a constraint.

A final reason I want to move to SoC/SBC/Linux is to get a head start 
with something that'll become ubiquitous sooner than you'd think today.

Already one can see a trend where such computers are being used where a 
BASIC Stamp, or a PIC microcontroller (the cheapest cost less than a 
dollar a piece) plus a couple of ICs and components (totalling, say $10) 
would be enough. But while the PIC processors have good facilities for 
debugging, downloading and ICE, the SoC/SBC/Linux offer ease of 
development, ease of firmware upgrades, in-situ major on-line 
adjustments, applicability of "standard know-how", a higher abstraction 
level, source code portability, vendor independence -- in other words, 
programmer productivity and time-to-market -- far bayond the (still 
admittedly) large cost difference per deployed unit of hardware, when 
production runs are not very long. And that's increasingly the case with 
small and medium size enterprises.

This situation is quickly becoming the norm instead of the exception.



More information about the Digitalmars-d mailing list