lack of contributors

[Brian Rice]

On Apr 5, 2006, at 12:12 PM, Waldemar Kornewald wrote:
> Brian Rice wrote:
>>> I just see that there are a few issues, but no syntax is perfect:
>>>
>>> (x < 5) /\ (x > 9) ifTrue: [...].
>>> vs
>>> (x < 5) \/ [x > 9] ifTrue: [...].
>>>
>>> and
>>>
>>> x < 5 ifTrue: [...].
>>> vs
>>> [x < 5] whileTrue: [...].
>>> ...
>> Brackets to denote lazy / repeatable evaluation is some kind of   
>> problem? Why?
>
> It's something you have to become aware of and it feels  
> inconsistent (though it's not). Maybe it's just me expecting the  
> language to be intelligent enough to know when lazy evaluation is  
> appropriate.

I'm not going to let the language source become ambiguous. Languages  
are notational technology, and precision is better than lack thereof.  
Languages should be dumb so that agents (human and automata) that  
work with it find it transparent. Again, see the AST MVC pattern.

> I want the language to help me, not stay in its static cage.

I want to be the King of all Siam and wear a fancy hat, but saying so  
won't do a damned thing to help that. I left the self-modifying  
concatenative syntax behind because it was too difficult to think in.  
Languages that try to be clever will bite you later.

> BTW, isn't that part of Zoku's ideas?

(Referring to http://www.zoku.com/ )

I don't find anything about Zoku to be terribly compelling other than  
the vaguely worded statements of philosophy. (Same with Paul Graham's  
Arc proposal.) It has lots of good marketing blurbs, but the  
technical content is dubious (I've seen deeper stuff than what's on  
the website), and I find the author to be somewhat of a quack. I have  
spoken to him in person many times (and also over Skype), and there  
is also controversy about his owning smalltalk.org and promoting his  
own personal views as even remotely resembling those of the larger  
Smalltalk community (or anyone at all, actually).

> I forgot to mention that if Slate should really support a  
> "scripting language" mode then it's probably good to support normal  
> text editors (UIs are not available very often). This would speak  
> for making Slate more readable even at the ASCII level (without  
> going too far away from Smalltalk's message syntax, but we have  
> discussed this already...).

See Common Lisp's SLIME mode with SWANK protocol. Someone could port  
this to Slate fairly easily if they focussed on it.

>> I will gratuitously comment that Slate can (right now) do:
>> if: i < 42 then: []
>> if: i < 42 then: [] else: []
>> and "[] until: []" / "[] while: []"
>
> I like this more than the current style because it sounds more like  
> a sentence.
>
> Loops should probably better look like this:
> while: [] do: []
> until: [] do: []

I understand, and these are not official yet, especially since I'd  
need to add bytecode compiler support to optimize them to the same  
level. Right now they're just layers atop the Smalltalk-80 idioms. In  
any case, "do:" is reserved for applicative settings where no other  
idiom suffices. I see no reason to ape other languages precisely when  
a close analogue is good enough.

In any case, WE CAN TRANSFORM THE SYNTAX LATER. It will also be  
easier with an IDE/source-database and pattern-matching template  
transformers for Slate code.

REPEATED NOTICE: Don't put the cart before the horse. Every single  
time you focus on these issues instead of handing me a patch, you  
irritate me. Doing the opposite will reverse this trend just as  
readily, and I exhort you to do so. If you do not do so, you will  
only raise my ire and accomplish less.

Is this too much to ask?

--
-Brian
http://tunes.org/~water/brice.vcf

-------------- next part --------------
A non-text attachment was scrubbed...
Name: PGP.sig
Type: application/pgp-signature
Size: 186 bytes
Desc: This is a digitally signed message part
Url : /archives/slate/attachments/20060405/a06f0cce/PGP.pgp