Hi Cortlandt,

after recovering from the state my evening plans from tuesday left me in, I 
can't help adding some more comments:

> Actually what I am talking about is a compiler with a grammar that reads 
> easily and the ability to create grammars and protocols easily as an 
> important feature.

As I said before, this added expressiveness comes at the tremendous cost of 
forcing developers to do something most of them normally  don't (and would have 
a hard time to do, I guess): Express their intends in a clear and unambiguous 
way without having an environment that forces them to do so by sheer lack of 
expressiveness.
I admit, though, that this point _might_ be made moot by an excellent 
development environment giving JIT code hinting in a way that makes it seem 
like it really enters into a dialogue with the developer.


About your (and Julian's) ideas about compilers (or VM's, really) that 
interpret a program's meaning by looking up lots of information on the internet:
I can only see two ways to implement this, one of which is interesting but not 
really that new and, first of all, very intricate, while the other one sounds 
like the holy grail of programming and is, like all holy grails, absolutely not 
feasible:
1. Build a huge library of common knowledge in a clearly defined way that can 
be consumed via some web service. This way, one could implement "tell jim to 
buy healthy food" by binding the term "healthy food" with the web service so 
that it's meaning is automatically derived from this gigantic pool of common 
knowledge. Ideally, the developer could choose whether a certain version of the 
definition should be used or whether his program should always use the current 
definition to benefit from, say, advances in human knowledge about what's 
healthy.
Of course, you'd have to have tremendous amounts of ressources to build up this 
library and some serious QA going to make this worthwhile. (For example, see 
the problems Wikipedia is going through, right now.)
This approach is certainly interesting, but, as I said, technically not that 
novel.

2. Automatically tap the collected knowledge of "the internet" by creating some 
web service that automatically creates the type of definitions described above 
on demand and from information gathered through search engines.
I think that this would be a huge task from a technical point of view, seeing 
as you'd need to parse and interpret free-form texts and whatnot. Also, I could 
see a whole new era of search engine spamming coming, with McDonald's trying to 
push a definition of "healthy food" that basically says "eat five quarter 
pounders a day, and you're fine".


What I could imagine would be some dialogue based development where you'd be 
able to have a conversation along the following lines:

Dave: HAL, I'd like you to tell jim to get some healthy food from the store.
HAL: Do you mean the store we talked about last week, dave?
Dave: Yes, HAL, that's the one.
HAL: Ok, Dave, I'll tell him. I'm gonna use the definition of "healthy food" 
you gave me two weeks ago, ok?
Dave: Yup, go for it.

and so on ...

But then again, would you really want to go to all that hassle only to - 
inevitably - get the dreaded "I'm afraid I can't do that, Dave" sooner or 
later? ;-)

cheers,
till

_______________________________________________
osflash mailing list
[email protected]
http://osflash.org/mailman/listinfo/osflash_osflash.org

Reply via email to