Re: What's right with FORTH and wrong with c.l.f.

Jonah Thomas wrote:
 > [...]
> But why go to that trouble when locals will put things wherever you want
> with no thought required? Similarly, why accept a simple string package
> that requires you to think, when you can have a complicated one that
> does everything you want behind your back and gives you results with no
> thought required?

Questions like this stem from an underlying assumption that all 
programming is the same.  Embedded systems, game programming, web 
programming, desktop applications, database applications, casual 
scripting, system tools, and so on... all the same.  Same needs, same 
priorities, same methodologies, same constraints, same same same.

Wrong wrong wrong.  Why accept a simple string package?  Because in some 
contexts, that's exactly what you need.  You need explicit and direct 
control.  You need efficiency.  You need a lack of abstraction.  Why 
then are there complicated string packages?  Because sometimes strings 
are not the problem you're trying to solve.  Because sometimes you're 
more than willing to put up with underlying complexity and implicit 
mechanisms because it makes expression of the algorithm clearer. 
Because sometimes you would prefer to be out playing with your kids than 
worrying about the efficiency of a string package that accounts for less 
than 1% of your program's runtime and memory footprint.

What I see in comp.lang.forth are people who come from a limited 
perspective-- for example embedded systems.  That's what they know. 
That's their world.  And so when someone who comes from a completely 
different perspective-- say someone who writes web applications using 
scripting languages-- the embedded systems people say "you should do
this."

And the opposite happens.  You see a handful of people here who think 
Forth is pointless and that high-level scripting languages are the way 
to go.  They show the same limited perspective.  Ask the guys who think 
Ruby is spiffy keen to run Ruby on a 8-bit microcontroller with 4k of 
ROM and 2k of RAM.  Ask the guys who rely on non-deterministic garbage 
collection in their languages to handle hard real-time requirements.

And yes, there is overlap.  I think my high-level application work 
benefits from an awareness of the underlying machine.  And I think my 
low-level systems work benefits from an awareness of higher-level 
abstractions.  Everyone can learn from each other.  But where there is 
no learning and no value is when conversations veer into absolutism and 
a denial that context matters.