Re: [pedantic-web] Re: [foaf-dev] [FOAF spec revised] foaf:logo aninverse-functional-property?

(excuse the x-post folks)

On Fri, Aug 13, 2010 at 4:04 PM, Jonathan Rees <jar@creativecommons.org> wrote:
> I was going to say that 200 responses to GETs of URIs that name members of
> foaf:Document were OK... guess not. Need to look for other examples.
>

So the issue (which I'll respond to properly on foaf-dev when I get
back from vacation) is essentially FRBR-related. FOAF currently has a
single very broad class 'Document' which encompasses all kinds of
docs, including the notion of 'that particular book in your hand with
the front page signed by the author'; but also a Web page. This
simplicity has served us pretty well, but can lead to confusion eg. in
the case of FOAF's sha1 property, which makes sense only when applied
to a specific bag of bits. Consider two different documents, both with
IDENTICAL content 'eg. the string "Nothing to see here.". They could
be written independently by different primary authors on different
days using different software, and yet have exactly the same sha1
checksum.

Until we get around to figuring out how to handle these radically
different notions of document which have to peacefully co-exist as
foaf:Document right now, we need a kind of modesty when we go around
saying what kinds of thing a Document can NEVER be. And since
foaf:Document covers real world artifacts, like a single physical
book, as well as sets of books (various FRBR views of literature), ...
 it is hard to sustain the view that physical documents can't ever be
directly identified with the physical artifacts that embody them. If
the spatial thing on my bookshelf can be a foaf:Document, so can a
parchment in a museum, a poster on a wall, a series of mime-typed
zeros-and-ones, the hardback edition of TimBL's Weaving the Web, a
signed copy of the same, a painting made on a wall in ink, or on a
body in paint...

All the time we only help ourselves to a single broad class for all
this diversity, we can't be too absolute in making disjointness
claims.

Much the same goes for Agent. Both FOAF and Dublin Core more or less
say an agent is something that does stuff (DC has slightly fancier
wording but equivalent sense). It turns out to be hard to define
classes of thing that can't be said to have 'done something' under any
circumstances; eg. the hailstone that broke the window, the storm that
flooded the field. Documents are similarly subtle, and the odd thing
for OO / DL style modelling is that these classes can still be
incredibly useful, even if they fall apart upon closer inspection.

Re FRBR-ish distinctions, my instinct is to avoid treating FRBR
concepts directly as classes, and instead to allow user-defined sets
to be more liberally used, but that's quite another story. Some rough
sketch of that here, ...
http://www.flickr.com/photos/danbri/2891150205/ and again more another
time!

cheers,

Dan

Received on Friday, 13 August 2010 14:29:29 UTC