info spaces

[cc www-archive]

<DanC_jam> sandro? that arch thing about info spaces was a fun read...

Thanks.  :-)

<DanC_jam> but the critical point, that the coffepot's address isn't a web page address, was made by assertion, with no supporting argument
* DanC_jam wanders off...

Yeah, I've been mulling over what to really say there.   I'll mull
some more here -- maybe you can suggest where I should take this.  

Like you and the TAG, I can't find any formally fatal problem with
having an object which is both a car and a web page.  Perhaps we can
agree that the range of the http URI scheme is "web pages"
(serializable collections of information), but we can't agree on what
that must be disjoint with.

I think the non-fatal flaw is that it doesn't follow the principles of
object-oriented design.  Objects and classes should obvious
associations with concepts in the minds of people familiar with the
domain of discourse.  You get this kind of muddled, nonsensically
overlapping classes in OO systems, sometimes, when people are using
multiple inheritence for code-reuse....

I dunno -- maybe it's not THAT muddled.  Maybe
http://www.drum.org/~natasha/pets/taiko.html is both a web page and a
dog, a "webdog", which I can both pet and gzip.   A dog with a web
page, AND a web page about a dog, ....

.....

I wonder if maybe some LSD would let me get my mind around that, cause
it doesn't seem to be happening otherwise.

[ If they're just saying they want a 404 http URI to denote Taiko,
then I think they're being silly.  If you're really never going to
have a web page there, then get off the web.  But the TAG has already
pronounced that one SHOULD provide content.  ]

....

Another modeling problem: 

What object (as an english noun phrase) is identified by each of these
URLs:

   (1)  http://www.w3.org/
   (2)  http://www.w3.org/Consortium

If they both identify the W3C, then how are they different?  When
should one use one URI instead of the other, ...?

Is each of them a WebPage+Consortium, where the consortium is the same
but the web page is different?

....

It's a nice idea, this use of a URI to identify things while having
the existing web infrastructure let you easily get information about
them....  RDF's leading approach (doc#term) has its problems too (like
colliding with XPointer); I wish we had a better answer.

....

[returning to the previous thread]

Maybe if the OO modeling rule were codified as an architectural
principle, then this conflation could be found in violation?

http://ootips.org/ood-principles.html
http://c2.com/cgi/wiki?OoDesignPrinciples
/// funny aside -- note the co-author:
The LiskovSubstitutionPrinciple (LSP): "An instance of a derived class
    should be able to replace any instance of its superclass"
[Liskov74] Barbara H. Liskov and Stephen N. Zilles, Programming with 
Abstract Data Types, Computation Structures Group, Memo No 99, MIT,
Project MAC, Cambridge Mass, 1974. (See also SIGPLAN notices, 9, 4,
pp. 50-59, Apr 1974.)
\\\

This modeling rule I'm proposing does not seem to be considered an OO
principle; I would guess because it's too obvious.   Maybe it's not an
OO principle -- maybe it's "common sense" or something.  

Hrrrm.  Actually, this is an ontology problem.  Until people express
their ontologies for web pages and for cars, any arguement about class
overlap is silly; after they create those ontologies, the degree of
possible overlap can be formally determined.  My quick sketch tells
me they'll come up as disjoint because information is not a physical
object and a car is, but if someone can construct an ontology which
allows overlap (because they defined cars or web pages differently
than I do), that ontology would probably help me understand the nature
of a webdog.    Developing an ontology of web pages would
be... interesting. 

    -- sandro

Received on Thursday, 5 December 2002 12:37:53 UTC