Thanks, I removed the older parts of the discussion since we seem to have reached a new state by now.
My comments are in the text of your last mail.

Ingo

-----Ursprüngliche Nachricht-----
Von: Mikael Nilsson [mailto:[log in to unmask]] 
Gesendet: Montag, 20. August 2007 11:02
An: Ingo Dahn
Cc: [log in to unmask]
Betreff: Re: AW: WG: Describing Application Profiles


lör 2007-08-18 klockan 11:58 +0200 skrev Ingo Dahn:
> Hi, Mikael,
> 
>  
> 
> thanks for your comments, perhaps we are converging, not sure yet. At
> least the differences in the views underlying our positions are
> emerging.

I think I agree, so let me summarize instead of commenting on all points
below.

1. The notion of APs: DC itself is about APs for metadata - i.e.
combining different kinds of metadata into one metadata record.

I do not question the usefulness of using metadata together with other
things (the IMS notion of AP) - I just wanted to make clear that the two
kinds of APs - let's call them intrinsic APs and extrinsic APs, have
*very* different functional requirements. And I'm also not saying that
IMS is doing that part (extrinsic APs) somehow "wrong".

Also, I certainly think that this group, as an example, *would* be
interested in seeing the DC-ED AP be used in such "extrinsic APs". But
we need to keep the concepts separated.

So, yes, I'm talking about the case where metadata + metadata =
metadata. Not metadata + XHTML = useful webpage, or XHTML + XHTML =
website. All useful, but not what I call intrinsic APs.

[ID] OK, I like this terminology.

Now that we've cleared that part, we can go on about "extrinsic APs". no
need to stop the discussion :-)

2. I also tend to find it problematic that some specifications do a lot
of metadata plus something else - i.e. the structural parts of IMS CP
are useful as metadata, but are too embedded in the non-metadata parts
to be easily combined with other metadata. I'd like to see a better
separation in specifications regarding what parts introduce new metadata
elements, and what parts are there for other purposes.

Also, the IMS specs could make good use of GRDDL

http://www.w3.org/TR/grddl/

i.e. as a way of extracting metadata out of "homegrown" XML languages.
I'd also like to see the IMS AP guidelines recommend this.

[ID] GRDDL is about transforming metadata from other bindings to RDF. As of today I am not yet convinced by the actual use of RDF outside controlled communities to invest efforts into such a translation. I know that it is possible to express everything in RDF (the same way as you can do first order logic without function symbols and with binary relations only, to refer to my beloved background ;-)), but I find it inconvenient, in particular when such transformations can be made available relatively easy when there is a real need for them.
I prefer a mix of structures and relations - to keep every information describing a single entity in a single XML element (like <lom>) and to use relations only when there are related entities.

3. The DCAM would not be well designed to handle "extrinsic APs" - it's
simply not designed to handle that at all.

[ID] Yes, DCAM is what it promises to be - description of the Dublin Core Abstract Model and its derivatives. Well it is probably suited to handle any abstract model following the property/value/vocabulary paradigm. It is not even metadata intrinsic (e.g. cannot handle LOM), but DC intrinsic. And it is not capable to handle _any_ binding which I consider as a significant deficiency.

4. Regarding LOM - I think LOM suffers from a lack of consideration for
the needs of "intrinsic APs" - it doesn't handle many of the
metadata-specific requirements very well. I think LOM at the core *is* a
metadata specification, only it's not very explicit regarding the
aboutness, etc. It's fixable, though.

For example "lom/educational/description/string" and
"lom/metametadata/identifier/entry" are not "about" the same resource.
This makes machine-merging of metadata difficult - it has te be handled
manually. So for "intrinsic APs", LOM has issues.

[ID] I don't think so. LOM is equally well suited for LOM-intrinsic APs as DC is for DC-intrinsic APs - everybody has his garden and his fence.

/Mikael