[PRL] values and expressions

[Matthias Felleisen]

In the #3 sense you have this:

Syntax: Expressions only. It makes _no_ sense to mention values here. 
This is often done to save space in the paper.

Semantics: Separate the set of expressions into a set of values and, 
implicitly, a set of computational constructs. Put differently, assume 
some magical way of getting a reduction relation Syntax x Syntax that 
specifies the syntax-based abstract machine where states are programs 
and execution steps are elements of the reduction relation. For such a 
machine, you must specify final states; call those values.  In a sense, 
this definition is implied but as the huge confusion of 
normal-order/applicative-order/normal-form of the last 30 years shows, 
it's better to spell it out. In our typical systems, this set of values 
also helps specifying the relation, which is why we do it first.

Plotkin did all this in 1972, set the applicative/normal-order people 
straight, and showed that leftmost-outermost is _the_ proper way to 
reduce. Period. Unfortunately, there are still confused people, but I 
have given up on those.

Now again, if you want to see weird scenarios, take a look at Crank and 
Felleisen.

-- Matthias

On Oct 23, 2005, at 3:02 PM, David Herman wrote:

> Sorry. I meant #3.
>
> Dave
>
> On Oct 23, 2005, at 2:14 PM, Matthias Felleisen wrote:
>
>>
>> On Oct 23, 2005, at 12:16 AM, Dave Herman wrote:
>>
>>
>>> Is it a really bad thing not to have the values of a language be a 
>>> subset of the expressions?
>>>
>>
>> Dave (et al), I must say that I am disappointed about this question 
>> (and some of the answers). As posed, it makes little sense:
>>
>> 1. Do you want to ask it in a denotational setting?
>>
>> 2. Do you want to ask it about an operational setting `a la 
>> Kahn-Milner?
>>
>> 3. How about a reduction semantics `a la Plotkin-Felleisen?
>>
>> The answer for each is different, because the word value means 
>> something entirely different. Nobody should have given you an answer 
>> without inquiring what you really want :-)
>>
>> -- Matthias
>>
>> P.S. My hunch is that Richard intuited your assumptions properly.
>>
>> P.P.S. For 3 with weird ideas, look at Crank & Felleisen some old 
>> POPL.
>>
>>
>