[Metadata] DOI for reference linking

Very interesting paper - the clarification of the need to identify the Work itself, as well as any Manifestations of it, is invaluable.

Manifestations, both physical and digital, and performances, may all have 'intermediate' pages to which a DOI can resolve (information, order form or whatever). That is in fact all that physical manifestations and performances can resolve to in the digital environment, although digital manifestations can also resolve to the object itself (although they will not do so, in the absence of an 'e-commerce in IP' infrastructure, unless it is available free of charge). Works, on the other hand, I don't think do have such intermediate pages since you cannot actually order/obtain a work.

All of the above do permit resolution of a DOI to a metadata record, and in fact this is likely to be an extremely useful resolution for all sorts of purposes and services.

In the case of journal article citations, one very likely (perhaps the most likely) scenario is that the citation link will first resolve to the article abstract. I don't think it is correct to see this simply as metadata (Description); it can be very much a tradeable entity in its own right, whether it is created by the author, the publisher or a third party (e.g. A&I service). The abstract then links to the full article, normally subject to specific commercial hurdles (e.g. 'are you a subscriber'). This may therefore need to be built into your model of how citation linking works (and this implies that the abstract will require its own DOI).

The perhaps unwelcome conclusion is that one journal article will need several DOIs with supporting metadata - work, physical manifestation(s) (print), physical manifestation(s) (CD-rom), digital manifestation(s), and the same again for the abstract. How to make this palatable to publishers is an interesting dilemma!

You refer interchangeably to 'Physical' and 'Print' manifestations. This is not strictly correct - a CD-rom is, in the INDECS model, a physical and NOT a digital manifestation.

Your diagrammatic classification of 'Type' would, I think, be better if Work, Manifestation and Performance were all on one logical level, and the alternatives Physical and Digital (for Work only) on the next level down.

Kernel elements for all DOI genres cannot, I think, make it normally mandatory to include an identifier from one or more other schemes. Only if it exists already - to create a new ISBN, ISSN or whatever purely for these purposes would be nonsensical, surely?

URL resolution cannot be mandatory - work does not resolve to a URL; physical manifestations, performances can only resolve to an intermediate URL.

Journal article kernel needs to state Form - Type may be Manifestation: Physical but this does not tell you if it's print or CD-rom, and this is necessary

Publication date may not be a meaningful concept when an article is first mounted in digital form - this may be before it has been allocated a 'place' in a journal issue (e.g. 'As soon as publishable' service).

'First (xxx) manifestation' might be better named 'primary' since the language does not then necessarily imply that it is chronologically first (for example, the 'first manifestation' of a journal article initially mounted in the 'As soon as publishable' service might, functionally, be the formally published one with pagination etc.)

You tend to refer throughout to the 'Publisher' assigning the DOI etc. While this is an understandable assumption for ex-publishers like you and me to make, it is not necessarily the case. Language needs to be scrupulously neutral to avoid alienating other players in the information chain.

One operational matter still concerns me. While I absolutely see the value of creating citation linking services as soon as possible, as a positive demonstration to all concerned that the DOI can deliver services of value to the information chain, I am unconvinced that this should properly be operated by or even under the auspices of DOI. Creating a standards and technical infrastructure to make it possible - yes, absolutely! But actually operating it - dangerous. I think this could be seen as monopolistic, or a least distinctly anti-competitive, and is more properly left to commercial players in the marketplace, of which there are plenty.