Case-sensitivity: why -- or why not? (was Re: Damnation!)

[Gareth McCaughan]

Tim Peters wrote:

> I haven't used Alice.  The counter-argument I've heard from some who have is
> that this was much more a problem with Alice than with Python, in that Alice
> exposed various predefined modules and classes with LongAttributeNames but
> in a variety of distinct styles:  case-sensitivity hurt because the graphics
> methods used case inconsistently in their names, and nobody could keep all
> its idiosyncracies straight.  With benefit of full ignorance, I have to say
> that sounds more plausible to me than that humans are baffled by
> *consistent* use of mixed case.  you find this sentence *inherently*
> jarring; i do too <wink>.
> 
>> ...
>> I am a very case-sensitive person myself: ...
> 
> DITTO.  Tools to help people get the case straight will be wonderful, but to
> the extent that case-insensitivity is an extraordinary claim, let's demand
> extraordinary evidence before building it into the language.  Didn't the ABC
> project run usability studies on this aspect too?  IIRC, ABC was also
> case-sensitive.

I very much agree with all this.

>> (The number two problem was 1/2 == 0; there was no
>> significalt number three problem.)
> 
> I'm wholly comfortable with 1/2 == 0 myself, but am equally comfortable with
> rationals, floats, constructives or BCD, and can easily see why 1/2 == 0 is
> a bad idea for computer newbies (more, it seems a bad idea for just about
> everyone except system programmers -- which we both were when this decision
> was made).  So I'm not just flaming Randy <wink>.

Here's a half-baked suggestion.

Give Python rational numbers of two flavours: with long
integers and with short (machine) integers. Make overflow
on "short rationals" produce floats rather than "long
rationals". Then:

  - 1/2 = (surprise, surprise) 1/2.
  - Doing lots of arithmetic using integers carelessly
    can't slow you down by an unbounded amount.
  - No one is going to be desperately surprised if
    adding two very large fractions yields a float.
  - We have proper rational numbers available for those
    who want them.

Of course, this doesn't fit well with something else I'd
suggest for Py3k (namely: make overflowing short integers
silently become long integers). Too bad.

-- 
Gareth McCaughan  Gareth.McCaughan at pobox.com
sig under construction