Re: [xml-dev] What is Data?

["Simon St.Laurent" <simonstl@simonstl.com>]

Peter Hunsberger wrote:
> That's essentially my second point below: the context for a human is
> necessarily different from that of any system operating on any data.
> Any attempt to attach a special status to any context is going to do
> so at the expense of some other context.  Pick your costs and benefits
> as best you can ...

Those differences necessarily make any discussion of data or information 
vague.  I clearly see a larger gap between the human and machine 
contexts than you do.

And if you'd rather not assign humans special status, I guess I have to 
break out Damien Conway's OSCON mantra of "What would John Connor Do?" 
to avoid living in the world you describe.  Costs, benefits - oh wait, 
who was deciding that?

>>> I'm actually being somewhat serious here.  I often try to get the
>>> people new to data modelling to think more "meta" by breaking it down
>>> as:
>>>
>>> - Information is data in context;
>>>
>>> - Data can be local or global, private or shared, transient or
>>> persistent, typed or untyped.  Anything that puts any constraints on
>>> any of these dimensions is some form of context.
>>>
>>> - One persons information is another persons data (substitute process
>>> for person as needed).
>>>
>>> One can also play with "Knowledge is information in context" and
>>> iterate, but that starts to devolve into debates on AI or metaphysics
>>> (depending on the audience) if you're not careful...
>> If your people think like computers, or in this case, prefer to think that
>> they do, this description is great.  If I spend too much time programming,
>> everything looks like a programming construct.  Step back, though, and it's
>> clear that I've just spent too much time talking with computers.
>>
> 
> Who said anything about computers? ;-)

I looked at the model of thinking you provided.  It felt a lot more like 
  the analysis we use in computers than the analysis people use in other 
contexts.  Your description of "data modeling" is basically "teaching 
people to think like computers."

That's useful, but I don't think it tells us anything meaningful about 
data or information.  (I should go back to the old rhetoric texts for 
contrasting models, I suppose.)

> Again, it's a question of what costs do you want to pay?
> 
> One can spend enormous amounts of time specifying a very precise set
> of contexts within which some system (automated or not) can handle
> some piece of data.  If one does so well, you end up with a system
> that can handle some very precisely modeled piece of data in a very
> efficient mater. That system, may however, have problems handling
> anything outside of that precise context (or set of contexts).
> 
> Alternately, one can allow for much more generalized contexts and
> spend more in terms or processing resources.  If one steps back too
> far, you're attempting to do AI, or metaphysics, or philosophy
> (depending on your context)....

Yep - there's no general solution.  None at all.  Never really has been, 
actually.

-- 
Simon St.Laurent
http://simonstl.com/