[sdiy] Secrets [Was: Discrete OTA]

[Neil Johnson]

Hi Paul,

> sorry, I disagree with both of you.
> Software development *does* cost money

This particular thread was started when Andre posted:

> The world of libre software is confronted to the same problems
> but has come to the opposite conclusion. They've been sharing
> code for decades. They never let arses and parasites stop them.
> There is now more code out there to read or hack than anyone can
> handle.
>
> If software hobbyists do it, why couldn't hardware hobbyists ?

In the context of "the world of libre software", then most definitely
the cost of starting up in software development is essentially the
cost of a PC, which is practically zero these days.

I further went on to say...

> The cost of entry of developing hardware is significant...
> ... to develop something significant ...  requires...
> ... maybe firmware development (it's harder than PC software),

So in that context, considering what you are up to, you are not
writing Software (capital -S) but what I would call firmware.  And as
I said, that *does* cost money to get set up.  So it seems you're
agreeing with my point :)

But just for fun, here are some more comments on yours:

> lets look at the cost of a compiler, a good one can be £2000 per seat.

I would say GCC is a good compiler (and for better or worse the
majority of Android phones on the market today have been built with
GCC), and the public version is free.  Of course, you don't get whizzy
IDEs with it, nor fancy CASE tools.  If you shop at CodeSourcery you
can buy a souped-up version with extra doodads.

And, quite honestly, £2000 isn't that much for an embedded compiler.
20 years ago I bought a copy of Keil C51 Professional Compiler for the
8051 8-bit micros (now _that_ was a damn fine compiler!) for not much
less than that figure - I see nowadays its about £2500 including VAT.

I don't dare ask the name of the compiler you bought.  I've used quite
a few over the years, and have in mind good ones and bad ones.

> So say you have two developers, there's £4K.

The compiler I use on a daily basis at work is costed at around £10k per seat.

> Next you'll need a couple of development platforms, say £1K a piece.

In the grand scheme of things that's peanuts.  The code I write runs
on custom processors on dedicated silicon.  The FPGA emulators I use
cost $250,000 each.  I have access to about 10 of them (they have
different configurations depending on what needs to be developed).
And we have more in other countries.

Your £1-2k gets lost in the noise.

> Now it's going to take time for the software developers to get up to
> speed on your platform and the task at hand, good programmers say two
> weeks.
> based on £10 an hour, two developers, 40 hours a week, there's another
> £1600.

Good programmers should cost a heck of a lot more than £10/hour.  If
you're only paying that little then you're not hiring good
programmers.

> Say the work takes 3 months, including integration and test, that's £9600.
>
> so, cost of your software development - £17200.

For a cash-strapped startup that might seem a lot, but in the wider
world, it's a pitifully small amount.

> remind me of how this is "zero cost" for software development.

As I said earlier, I think you missed the start of this thread so
missed out the context.  For "libre software" the startup costs can be
zero.  For embedded firmware there are real costs due to the physical
nature of the task.

> Software isn't zero cost, but it has a much lower "direct" cost.

Done right I would put firmware development costs at about even with
purely hardware development costs.  By the time you've accounted for
JTAG boxes, programmers, logic analysers, protocol analysers, maybe an
ICE (do companies still make proper ICEs with bondout chips these
days?) ) if you need to get that deep, and so on.

> Where software does score is that to reproduce it requires almost zero
> cost.

Indeed, although in an embedded product it might be sitting inside an
EPROM or FLASH.

Neil
--
http://www.njohnson.co.uk