Document Design and metadata

[Patrick Durusau <patrick@durusau.net>]

Savante,

I have been thinking about your position that there should be no 
indirection of metadata by having metadata on metadata in a document and 
think I can articulate the difference in our positions. See what you 
think about the following:

The elimination of indirection of metadata, that is that any metadata is 
applied directly to the content of the document (and by implication not 
to metadata) is a document design issue. That is to say that from the 
standpoint of designing the document format that whatever metadata is to 
be applied, it should be applied by appropriate mechanisms to the 
content of the document.

While attractive and perhaps even good design, ;-), that works if and 
only if my only concern is designing of the "best" document format. 
Unfortunately, we know in fact that our document design is going to have 
to accomodate converted documents that followed less than the "best" 
design and even when they did, users will have in fact abused those 
designs.

So to me the issue is whether the no indirection position can accomodate 
the conversion of other document designs and/or user abuse of those 
designs.

For example, in our last call we discussed the need to have metadata on 
metadata that could change the processing of and ODF document. Take 
sections, which I understand cannot be embedded in other sections in 
OOXML (disclaimer, I have not verified that statement for myself). In 
order to maintain a roundtrip, I have to communicate to the ODF 
application that it cannot embed sections within sections. One way to do 
that would be to add metadata to the section style representing that 
prohibition. Actually as I typed that example it occurred to me that 
such metadata would be better attached to the document itself to prevent 
the creation of other sections that would embed in sections. (Whether 
than can be generalized for all other round tripping questions I don't 
know.)

Hmmm, well, take a user abuse senario. When I use "italic" style, it is 
solely to mark foreign words in a text. Other people use "italic" for 
emphasis on the introduction of a new term in the document. Downstream, 
that is outside of the author's control, how are we to apply that 
directly as metadata to the content? (Assuming such a mechanism is in 
fact available?) If I can add metadata to styles, I could mark my 
"italic" style as being applied only to foreign words, which would allow 
any spell checker to simply skip those words.

The same question comes up with Rob's example of embedded quotations 
where we want the spell checker to skip those quotations.

I suppose you could say that we can extend styles to carry with them 
information about invocation of the spell checker, so that there is no 
indirection but that is only one example and I am not sure that we would 
be able to account for all the cases.

BTW, does it matter who applied the style? Say for example that I apply 
a style that "hides" part of the document. Is it metadata about the 
style as to who applied it?

Summary, note that I don't object to the notion that avoiding 
indirection is a good design principle but my concern is that there are 
lots of documents that have not followed good design and/or have been 
abused by users.

If, for example, the any markup carried with it a "source" attribute, 
which indicates the origin of the markup, say OOXML, then an ODF 
application that was to be used for round-tripping could enforce the 
appropriate behavior to insure the document could be round tripped. (And 
no, I don't think we can cherry pick the appropriate elements for such 
an attribute. Simply not possible to know all the possible combinations. 
Applications could simply choose to ignore the attribute.)

Hopefully this captures some of the issues underlying the question of 
indirection of metadata.

Hope you are at the start of a great week!

Patrick

-- 
Patrick Durusau
Patrick@Durusau.net
Chair, V1 - Text Processing: Office and Publishing Systems Interface
Co-Editor, ISO 13250, Topic Maps -- Reference Model
Member, Text Encoding Initiative Board of Directors, 2003-2005

Topic Maps: Human, not artificial, intelligence at work!