W3C home > Mailing lists > Public > www-tag@w3.org > December 2002

Re: WebArch Ambiguity about Objects, PLUS Suggested Major Replacement

From: Sandro Hawke <sandro@w3.org>
Date: Mon, 30 Dec 2002 21:02:23 -0500
Message-Id: <200212310202.gBV22Nu14229@wadimousa.hawke.org>
To: Mark Baker <distobj@acm.org>
cc: www-tag@w3.org


> >   The owners of Mark Baker SHOULD make available representations
> >   that describe his nature and purpose.
> > 
> > Obviously, Mark Baker doesn't have an owner, so this is nonsense.
> 
> No, it seems fine to me.  I own me.

Perhaps, perhaps not.  Clearly no one owns the sun and moon, the
English language, the weather in Los Angeles, or the latest events in
the world of technology.  When people talk about each resource having
an owner, obviously they are thinking of a limitted set of possible
resources.  I believe they are usually thinking about knowledge bases
/ web pages / web sites / documents / whatever you want the call the
maintained collections of information that HTTP URIs denote.

> That perhaps a bit of a twisted example, but a more reasonable one
> would be;
> 
>   The owners of Dan Connolly's car SHOULD make available representations
>   that describe its nature and purpose.

Everyone should make a web page about their car?  Huh?  No.  The idea
here is that everyone who publishes a web address (even if they are
only using it to identify something like a media type) SHOULD make
sure it is properly served.

> > > The Web is a universe of resources. 
> > 
> >   The Web is a universe of things like Mark Baker, his house, and
> >   his coffee pot.    
> > 
> > Somehow this is different from the real universe?   Saying the web is
> > the universe is nonsense.
> 
> "is" is too strong; I wouldn't say that.  But I have no problem saying
> that the Web includes everything in the universe.  It sounds wacky,

Excuse me, but it is wacky.  :-) The universe is the only thing that
includes everything in the universe (by a common definition of the
word, at least), so if the web does too, then it is the universe.

> sure, but try to identify something in the universe that *can't* be
> identified by a URI, nor return a representation on a GET (even if it's
> not an http URI).

You're just playing the object-oriented mind trick on yourself!  Just
because you can model everything in the universe in your computer
system, don't think your computer system *is* the universe....

> > The distinction I'm trying to make only matters when information
> > systems collide.  For example:
> > 
> > Within a particular application, perhaps a school's class-registration
> > database, students can use their home page URLs to identify
> > themselves.  That works fine until that data gets merged on the
> > Semantic Web with data about, oh, student's home pages.  Now we're
> > saying things like "there is a thing enrolled in CompSci 101 which has
> > a last-modify date of 15 minutes ago."  This is really not a very good
> > information system design!  On the other hand, it's just fine (and
> > really not much more work) to say "there is a thing enrolled in
> > CompSci 101 which has a home page which has a last-modify date of 15
> > minutes ago."  Wouldn't you prefer we keep these things straight?
> 
> Absolutely, but I'd say we already can/are.  If we know that
> "http://www.markbaker.ca" identifies me, then it is incorrect to assert
> that the resource identified by "http://www.markbaker.ca" was "modified
> 15 minutes ago".  So we need another way of talking about the
> representations.

By "representation" here you seem to mean the data stored in the web
server.  Conceptually that's not the same as the external data
representation the server provides in response to a GET (which is what
WebArch calls a "representation"), even if they are bit-for-bit the
same.

>  With this example, I assert that
> "http://www.markbaker.ca/index.html" identifies the HTML representation
> of me, so it's quite reasonable to make a last-modified assertion about
> it.  As Roy mentioned[1], another means of talking about representations
> may also be useful.

I assume you mean that "http://www.markbaker.ca/index.html" identifies
a collection of information about you (which happens to be stored in
HTML and probably English).  And that's what people get when they GET
"http://www.markbaker.ca".  And it's also what people get when they
GET "http://www.markbaker.ca/index.html".  But if you change the data
on your system, the last modify time on
"http://www.markbaker.ca/index.html" changes, but not last modify time
on "http://www.markbaker.ca".

If you really think this is a good design, you're probably right that
there's not much point to you and I continuing this conversation.

It seems to me that the web makes a great deal more sense when one
thinks of http URIs identifying maintained collections of information
(which TimBL just calls "documents").  Is there anything particularly
novel, bizarre, or lacking in such an approach?

> BTW, I don't think we're covering any new territory here from the
> httpRange-14 discussion, so if you want to continue, I suggest we take
> it to www-talk.  Followups set accordingly.

Forgive me, but it seems to me that you're not understanding me, so
you're in no position to judge whether what I'm saying has been said
before.  Of course I may have missed some bit of discussion in the
past; feel free to embarrass me :-) with some pointers to messages
which specifically answer my points.

    -- sandro
Received on Monday, 30 December 2002 21:06:12 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Thursday, 26 April 2012 12:47:14 GMT