[Oberon] Commercial system
lab.eas at gmail.com
Sat Feb 8 18:41:47 CET 2014
Evolution vs revolution.
Is the X86 instruction set 'superior'?
Or is it like 'english': just got established first by the luck of history?
IMO WinTel is technically a mess, because of the commercial need to evolve on
their sub-optimal historical path. But therefore a great commercial success;
until a competitor, who is free from the baggage of history, beats them.
Will ARM-android over run WinTel?
OTOH the evolutionary/one-step-at-a-time vs the optimal-integrated-design
method is often very cost effective for small productions. And with a chaotic
world: small production runs are common.
Eg. with *nix piping you can do:
= show me the location of the utility: M
= use its dir with file: *3W*
= look in that file & get the 2nd reference to "FC1"
= call THAT executable with <the argument>.
*nix jockeys write this all in one-line!
But my point is that each of the 4 steps/transformations takes you closer
to your goal. And most importantly, if it's OK up to stage N, you never
need to go-back, nor even remember/know how you got there. There are no
names to remember. There's only "it".
Since it's your time/effort that counts/costs and not the hardware, this
compositional style has become cost-effective for advanced computer users.
BTW, since ETHO/LEO can call linux executables, it can nicely handle
Reading the HOW2/instructions eg. of how to get-on-line via a USB-G3-dongle
one sees that the writer doesn't appreciate the nice, extreme-modularity of
the system. Once the LED lights, the preceding 6 stages have 'succeeded'.
Ie. it should be seen as N individually testable stages/modules.
Has ETHZ's FPGA device got any debugging LEDs [like rPi], or will the victim
need an oscilloscope...etc, to confirm?
If you can't put a simple block of debugging bytes at the start-vector adr,
for debugging/confirming purposes it's no good for teaching.
And the hardware must be testable at a low level, independent of Oberon.
The powersupply voltage is decoupled from the semicolon in the language syntax
It's not a magic-box that eventually comes to life after sufficient effort.
It's N-stages, [although not nicely linear like the above example] which
should be as far as possible, independently testable.
Some one should write the confirmation/test-stages for Gmail via pop,
instead of assuming it's a flight-to-Mars, where you can't test incrementally.
== Chris Glur.
More information about the Oberon