Re: [Ref-Links] Re[2]: [Ref-Links] DOIs used for reference linking

[Mark Doyle <doyle@aps.org>]

Greetings Ed,

On Fri, 2 Apr 1999, epentz@harcourtbrace.com (Ed Pentz) wrote:

Thank you for your reply. I did look at the IDEAL links page when preparing  
my paper (and I admire it because it includes extra functionality which I  
haven't added to the APS link manager, for instance for locating tocs).

> The APS Link Manager is very similiar to the IDEAL Links
> system.  While we allow, amongst other things, traditional
> citation data (ISSN, volume, first page) to be used for links
> to IDEAL, we also accept DOIs (see attached idealink.pdf).
> Using DOIs or traditional citation data isn't an either/or
> situation.

That of course is true. The question is what new thing does a DOI based  
solution bring to the table.

> With reference links there will always be two
> things - the traditional citation plus a URL (either a DOI or
> an "intelligent" code or a DOI made up of an intelligent
> code).

Agreed.

> The main drawbacks I see with Mark's distributed, algorithmic
> (my label) linking proposal are:

Good label....

> *every linker needs to know the algorithm of every linkee -
> this means work is distributed and the algorithm metadata
> must be distributed.  (non-trivial for the likes of ISI and
> for publishers with a multidisciplinary set of journals).
> This is really how things happen now and there is agreement
> it's not scaleable.

This I don't agree with. While it is true that the linkers do need to be  
able to know aobut each other and share algorithms, I think a lightweight  
solution like Eric Hellman's SLinkS can provide a simple, centralized way to  
share the necessary metadata (CODENs/ISSNs and volumes/years online, URL  
template based on traditional metadata). Once you have a publisher's  
template, you can construct all of your links dynamically. Checking for  
validity, if desired could also be done, but this is a logically independent  
step. I think this is more scalable than every publisher in the world  
contributing to a single database that must be maintained and updated  
(perhaps daily). Publishers locking themselves into a single system is quite  
dangerous. In addition there is the question of who should control the  
centralized server.

> *the linker needs to take a journal title abbreviation from
> a citation, identify the publisher and then apply the
> algorithm (including any non-standard journal abbreviation,
> ISSN or Coden)

Right, we do this cleanup/tagging anyway as part of our production process.  
Personally, I would like to see the burden of this step put more onto the  
authors by having their citations checked and linked at submission time, but  
I digress. This step is logically independent of the DOI intiative or any  
other linking solution.

> *the linker has no way of knowing if the cited article is
> available online unless they have metadata about the online
> availability of the journals. (at the moment, even with the
> IDEAL Links system, we have to send article metadata to
> those linking to IDEAL so they know what's there and get
> ISSNs)

Alternatively, something lightweight like SLiknkS could provide the  
necessary metadata which is a much smaller data set than the metadata for  
every article. Publishers don't put articles online willy-nilly, but in  
well-defined blocks (for instance starting with a particular volume or  
year).

> Some things need to happen before the DOI system can work as
> a scaleable solution, but the idea is that publishers will
> take citations and batch submit them to a central service to
> be matched to DOIs.  This requires a metadata database and a
> matching service - both seen as a critical applications by
> IDF members.  In this scenario the publisher, or linker,
> doesn't need to know anything about the cited journal or its
> publisher.  It could be very easy to submit citations and
> get back DOIs that can be used for linking.

This is an important service, but fundamentally separate from the linking  
phase. Once a citation is validated, then if you have the algortihm, you can  
apply it dynamically without storing any additional information. A  
competing proposal would be PubRef which could provide the service, but it  
wouldn't require the DOI part. Or Sompel's SFX server. DOI's have the effect  
of having to require keeping around the DOI in a database (since most are  
not algorithmically derivable it would seem) and adding an additional  
indirection into the loop (any publisher who uses URL's in the DOI service  
that don't also incorporate indirection is asking for a maintenance  
nightmare).

> With the DOI there is nothing stopping a publisher
> incorporating their algorithmic identifiers into a DOI.
> Academic Press is using is own article identifier as the
> DOI, which means that the DOI is integrated with our
> internal databases and systems.  We found it fairly easy to
> implement DOIs and create metadata incorporating those DOIs.

Except that this seems to violate all of the DOI rhetoric about not  
creating intelligent identifiers -- if you are going to add intelligence to  
DOI's, then you are removing some of their nominal advantages. In any case,  
if publishers start doing this, then you can algorithmically create the  
DOI's, but then, why not go directly to a publisher's linking service? You  
would still need to know how a publisher encodes their metadata in the DOI,  
again requiring the need for an SLinkS like service (perhaps as part of the  
DOI server).

It would seem that publishers are all converging on implementing link  
manager like solutions (APS, AIP, Academic Press, IOP, Elsevier, Springer,  
EDP Sciences, Japan Journal of Applied Physics, and others have all  
implemented or are actively developing them). Incorporating them into an  
SLinkS type of service with the necesary metadata would allow for robust,  
scalable linking. Validation of citations during production is a separate  
issue that may require a centralized repository of metadata, but this is  
fundamentally a separate issue logically independent of the DOI initiative.

Best Regards,
Mark
The American Physical Society

------------------------------------------------------
Ref-Links maillist  -  Ref-Links@doi.org
http://www.doi.org/mailman/listinfo/ref-links