Re: Apologies to gnome-gui if I ever said you guys were flamers...

[Tom Vogt <tom lemuria org>]

Dan Effugas Kaminsky <effugas@best.com> wrote:
> Could somebody please tell me what technical issues I have ignored?  When I
> was challenged on technical grounds re: screenplays, I started communicating
> with the author of Xlab and posted his things here.

this is actually a very good example.
your whole proposal of the screenplays was ignorant of the fact that xlab
existed. you found out that 90% of the functionality are already there only
after someone on this list pointed its existence out to you.


> one.  Cluehunting?  I fail to see how "it's been done before" and "it's not
> possible" are NOT mutually exclusive.

but you have ZERO information on HOW it could be done.

here's another example for you: that game I'm working on at the moment
employs a simple, but functional natural language parser for input and a
similiar simple, but working natural language generator for output.

I started about the same way you did, by thinking "wouldn't it be cool if
the program communicated with the user in the user's language?". but I
didn't go around and demand webspace for this idea. and en route to
implementing it I found out that both fields - natural language parsers
(nlp) and natural language generators (nlg) are VAST fields of current
research. almost without knowing it, I've moved to the very frontier of
current date computer science. that's when I backed down and had to REFINE
my proposal to make it feasable, to make it possible to get both the result
I was looking for and still be able to come up with it before 2020.

the result is a parser that will take english sentences and correctly
interpret them, to the level of understanding various qualifiers - e.g. as
the target of your action you don't have to write the name of the other
player, you can identify him by virtually anything that's unique. it also
will recognize what you mean if you made spelling mystakes :) or typos.
however, anyone seriously working in the field of nlp's will break down
laughing when I show him the underlying engine.
the nl-generator works in a similiar way.

now, to return to the point after this long example - what people are
missing is this refinement step. and idea is a nice thing, but you have to
do more than that. you've got to add some research into the topic, look if
similiar things are around, how they work, what their advantages and
shortcomings are. you have to get at least enough technical know-how to find
out if your idea is possible at all to code, if it will be hard or easy, and
where it will interfere with or enhance other parts of the whole project.
and then you have to sit down and rework the idea. I wouldn't even call it a
proposal before you did that.

anything else - and I mean ANYTHING, no matter how good as an idea - is just
sand in the machinery.



-- 
Those who do not understand Unix are condemned to reinvent it, poorly.
		-- Henry Spencer