[plt-scheme] Teaching Scheme

On Tue, May 04, 2010 at 04:35:44AM +1200, Samuel Williams wrote:
> Dear Shriram,
> 
> Thanks for your reply.
> 
> On 4/05/2010, at 2:31 AM, Shriram Krishnamurthi wrote:
> 
> > Let me caricature (a part of) this exchange:
> > 
> > Samuel: ... interpreted vs compiled ...
> > 
> > Shriram: nonsensical!
> > 
> > Samuel: well, most X's have interpreters, most Y's have compilers, so
> > it's reasonable to call X interpreted and Y compiled
> > 
> > Doug: convention is good, and anyway some languages are more static
> > and others are more dynamic, so we can use "interpreted" and
> > "compiled" as proxies
> > 
> > Samuel: yes, the term doesn't quite make sense (think about JITs), but
> > also, what Doug said!
> > 
> > To which I offer several responses:
> > 
> > 1. Notice how you (Samuel) subtly shifted your argument.
> > 
> 
> Thanks for your clarification. I feel that it is simply a matter of being aware that there are a vast range of terminologies and that not everyone agrees on the same meaning. I think its perfectly valid to be aware that terminology can be used in different ways with different meanings, and I appreciate your insights.
> 
> I don't think I've ever made an argument one way or another - I have simply done my best to use accepted conventions - but I appreciate that this might not be the best approach...
> 
> What is really important, as you have pointed out (point #3), is the general biases and assumptions made by the affiliation of a particular terminology to a particular language, and in a sense its implementation. I think this is actually a very important point which I haven't considered much.
> 
> > 2. If you want to use "interpreted" and "compiled" as aliases for
> > "dynamic" and "static", why not just use "dynamic" and "static"
> > instead, instead of abusing terminology?  Indeed, if you are trying to
> > create an EDUCATIONAL site, why not EDUCATE people that their
> > terminology is wrong, rater than propagate their mistakes?
> 
> This makes a lot of sense. So, do you separate it like this:
> 
> > 
> > 3. These terms are not value-neutral.  People hear "compiled" and
> > think "fast"; they hear "interpreted" and think "slow".  (Admit it, so
> > do you.)  Of course, this might have nothing to do with the reality of
> > implementations.  But it damns languages prematurely.
> 
> Well, in a sense, it depends on terminology and understanding. When I think compiled I think rigid and structured, and when I think interpreted I think potentially slow and somewhat more flexible.
> 
> I think generally this is a relatively well accepted understanding - but I'm not suggesting that it is correct. From your arguments I believe that it might not be a good way to portray languages.
> 
> > 
> > 4. Sometimes, these terms are references not to the implementation
> > strategy but to the interaction interface: "interpreted" means "offers
> > a REPL" while "compiled" means "works off a command-line".  But in
> > fact that is another bogus alias.  Furthermore, it does a great
> > disservice to compiler writers for languages like Scheme, who've taken
> > the trouble to write actual compilers while providing the *interface*
> > of an "interpreter".  Why don't you give them credit for their work?
> > 
> 
> Sure, this makes sense.
> 
> > Personally, I learned Scheme using a compiler and C using an
> > interpreter (really), so these terms have always struck me as bogus.
> > 
> 
> Understandably so.
> 
> > As for paradigms, I've made my point in my document.  Glad you enjoyed
> > reading it.  If you insist on using absurd terminology, I won't get in
> > your way.
> 
> Well, while I think your response is overly harsh, I think you make valid points and I'll certainly do my best to improve the standard of the material.
> 
> Can you give me some feedback about the categorization:
> 
> Statically Typed:
> 	C, C++, Haskell, Java, Pascal

Modula 3 -- a little-known systems language that's a lot better than a 
lot of more popular languages

> 
> Dynamically Typed:
> 	Ruby, Python, Scheme, Smalltalk, Perl, Basic, JavaScript, PHP

I've found that the most important distinctions between protramming 
languages are:

(1) Whether the langauge is secure.  i.e., whether a rpogram can break 
the run-time model and enter into the realm of indescribable chaos?
(If it's secure, its dynamic storage pretty well has to be 
garbage-collected)

(2) The type system.  Is it static? Is it dynamic?  (In my opinion, the 
word "type" should only be used for static type systems; if it's 
dynamic, the so-called 'type' is actually just part of the data.  Most 
people don't use the terminology this way, though)

(3) What types are supported, to what extent, and can users make more of 
them?

-- hendrik