[Prev][Next][Index][Thread]

Re: A question

> Well, I would describe Dylan as "struggling for traction". I am hopeful
> that this fine language will see commercial success with FO (the first
> time that there has been a company whose sole interest is Dylan) and
> that the Gwydion open-source effort will pass critical mass. I am
> playing with FO Dylan in spare moments, off and on, to see what is there
> now.

It is a real shame too, as next to the language "Self" (which fizzled
out and I have no idea why as it was *so* cool), Dylan is the only other
language I am aware of which even approaches a 21st century language.

I can only hope the FO guys succeed.

> As for Smalltalk, single inheritance is just flat out a mistake in
> language design. (If you disagree, don't bother trying to convince me,
> because you won't create even the tiniest bit of doubt in my mind.) I
> cannot stand the thought of learning another of those SI class libraries
> with an outrageously deep hierarchy with functionality pushed up into
> totally counterintuitive classes in order to get it where it can be
> shared. A well-designed MI-based framework/library is so much easier to
> understand and extend than an SI one.

I agree 100%.

> Of course I'm not trying to claim that you should choose a development
> tool based solely on the SI/MI dichotomy. I just wanted to point out
> that shortcoming, which I personally think deserves more weight than
> most give it.
> 
> BTW - Smalltalk can be forgiven this flaw because it was pioneering many
> concepts when it was created; certainly a couple of decades of
> experience with OO were bound to come up with some additional valuable
> concepts. Java, OTOH, simply has no excuse. Sun lazily left out a key OO
> feature and now invests many millions in convincing developers that MI
> is not in Java because it is "bad" and leaving it out somehow improves
> the language. Purest bullshit.

... a real cop out by Sun.  I expected more from them, but I guess
that's where big bureaucracy leads.

> And there are other Dylan goodies: for instance the macro system and
> multimethods.
> 
> Hey, don't even ask me what multimethods are good for. At this point I
> think my understanding of them is analogous to a (perceptive) neophyte
> who has just been "taught" about recursion with the infamous factorial
> example. It's a neat language feature that I'm convinced must be very
> powerful, but I don't quite know what to do with it!

Multi-methods are also one of it's best features.  Eg, if you have an
object and you want it to display itself in an arbitrary windowing
system, multi-methods will dispatch you to the right function.  So,

	      display(listBox, motifWindow)

can dispatch on both listBox and motifWindow to call the function which
is specific to both classes.

> Next time pick something easier. Ask us to help you choose between Dylan
> and C++, or Java, or Objective-C, or Object Pascal, or Cool ;-)

. I understand, Harliquin's Dylan provides incremantal compilation (in a
  running application) though I have not used it.  I would like to know
  how well it works.  Is Smalltalk interpreted?

. Dylan's container classes are second to none with a rich extensible
  hierarchy.

. Excellent exception handling (though I know nothing of Smalltalk's)

. Class and Subclass slots as well as virtual slots.

. Flexible typing system allowing easy prototyping up front, and
  performance tuning later.

> "Stephen J. Guthrie" wrote:
> > 
> > I purchased the Dylan Professional from Harlequin about a year ago and
> > stopped using it entirely after the recent events with FO.  I'm considering
> > picking the language back up but I would like an objective evaluation as to
> > why I should.
> > 
> > I recently began programming in Squeak (Smalltalk).  Give me a reason to
> > come back to Dylan.
>
Follow-Ups: References: