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

[Mark Doyle <doyle@aps.org>]

Greetings Larry,

On Thu, 08 Apr 1999, Larry Lannom wrote:

Thanks for your comments.

> To say that DOIs are not likely to appear in citations, which I think is
> what Mark Doyle said, and then to argue from that point that they would
> be superfluous to reference linking seems a bit tautological. You can
> make different assumptions and reach different conclusions. Recall, as
> an example, that the initial object of the DOI was not reference linking
> but intellectual property. So assume, for the sake of argument, that the
> DOI exists for reasons outside of reference linking. Does it then make
> sense to also use it for reference linking?

My point is that researchers (authors in general) are an undisciplined  
bunch (having been one myself and from my experience helping run the Los  
Alamos e-Print server and now working at the APS). Expecting them to provide  
and make regular use of long, dumb identifiers for citing their work, even  
alongside the usual citations, is unrealistic. That is my starting point.  
Lecturers won't flash them on their viewgraphs, people won't refer to them  
in conversation, and they won't have easy mneumonics for remembering them.  
Even with something like the e-print identifiers which are highly mneumonic,  
flashed in talks, and bandied about, researchers are quite varied in their  
willingness to provide dual citation information, despite that fact that  
doing so makes it trivially possible to hyperlink their articles to their  
citations. Thus, we are battling a long history of author behavior and this  
problem needs to be tackled before DOIs can become ubiquitous. The burden  
falls on the publisher to add the extra information and track it. In part I  
am questioning whether this burden (which immediately scales to include all  
publishers who want to provide linking) is worth it.

As mentioned as an aside in another message, I think that one solution to  
this problem is that when authors electronically submit their work, their  
citations and their linking would be automagically checked and the author  
would be given immediate feedback so that errors can be fixed. That way the  
burden is spread across the large number of authors instead of just the  
production process. But it is still questionable whether it is worth having  
to do the extra step of carrying DOIs around, forever and whether authors  
will in fact accept even a small part of the burden.

So I see this a barrier to acceptance and wide use, not a tautology. Your  
point about DOIs existing a priori is a good one though. But the trend so  
far is for publishers to do their own access control and I am not sure how  
the intellectual property aspects of DOIs applies to the narrow scope of  
scholarly publishing.

Another practical issue is that a Work (its abstract, its PDF version, its  
XML version, its paper version) may map to many DOIs - which DOI does the  
author supply? It would seem that decisions about what is to be linked have  
to be made very early in the process so that the right DOI is tracked.  
Changing that decision could mean updating many, many DOIs (at least under  
the current proposal).

> I agree with Eric Hellman that the tough parts of this problem are
> organizational and not technical. This is true, in my experience, across
> most of the digital library problem set.

Agreed! Publishers have only minimal control over the behaviour of the  
research community. Publishers should be able to excerise more control over  
themselves and agree to the use of standards for linking that factors out  
the need for the research community as a whole to adopt them.

> I also think it would be useful, to the degree possible, to both push
> the scenarios a few years into the future and pop the discussion up a
> level or two. Anyone creating, maintaining, and making available on the
> net large numbers of digital entities is going to have to have some kind
> of coherent information management scheme in place. Given a decent set
> of tools for accomplishing that, then I would say that the difference
> between updating a few algorithms and updating a few hundred thousand
> identifiers is zero, or at least transparent to the maintainer.

I don't believe that will be true. There is a lot of tracking of  
information that has to flow through an entire process and well-designed  
processes in principle should make it tranparent. But data integrity is  
going to be extremely important here and I think it is non-trivial to  
establish and maintain the necessary flow of information from beginning to  
end. Updating the centralized server (and all of its mirrors) for hundreds  
of thousands of records won't ever be trivial.

> The more
> interesting technical issue at that point, I believe, will be to find
> the best way to make that information accessible in an ad hoc fashion to
> whoever needs it. Herbert Van de Sompel's paper on SFX, an updated
> version of which will appear in D-Lib this month, provides a useful
> perspective on all of this. But again, the organizational issues are
> paramount.

Yes, this will be an important aspect of any solution.

> Finally, it may be a little misleading to always think in terms of large
> publishers. We publish a single electronic journal and assign 4 or 5
> identifiers a month. Given the vanishingly small barriers to entry for
> electronic publishing, it seems to me that there will be a lot of this
> in the future (although most people reading this list probably have a
> better idea of this than I do.

Perhaps, but the costs involved and the scaling issues only become apparent  
on the large scale. And acceptance by the large publishers will probably be  
a necessary condition for the success of the proposal.

Regards,
Mark

Mark Doyle
Research and Development
The American Physical Society


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