[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
[Ref-Links] IDEAL Links
Guenther:
Re this para:
"The other type of link is more complicated, since we have to
build the address according to the publishers directory structure.
http://www.idealibrary.com/cgi-bin/links/citation/0019-1035/130/323
If the publisher decides to change the directory structure, we
would need to rebuild our link files. This is no big deal, since
these files are created through scripts. However, it would make
it easier if all publishers would provide name resolvers for
their on-line data. Whether these resolvers accept bibcodes,
DOI's, or other identifiers (or preferably both) doesn't really
matter for me, since I have to keep the concordance anyhow."
Just for the record, IDEAL does not map links directly onto a directory
structure, so you will not need to rebuild your link files. And in fact the link
to Icarus you cited is similar to the U. of Chicago Press link described, only
using less metadata, ie just ISSN, Volume number and Page number - the standard
citation triplet (no year etc required - note that ISSN was chosen over Journal
abbreviation for robustness). The two following links are identical (the
cgi-bin/ in the pathname is optional):
http://www.idealibrary.com/links/citation/0019-1035/130/323
http://www.idealibrary.com/links/doi/10.1006/icar.1997.5822
So you can link to an article on IDEAL either by its citation metdata or by its
public identifier (DOI). Both are equivalent, although in the end the identifier
is guaranteed to be unique. With metadata there is always the possibility of
collision, ie two articles starting on the same page (which could be avoided by
enhancing the metadata set and so the complexity). The identifier is always to
be preferred, although the metadata route is generally more user friendly.
Metadata has a tendency to be too much or too little - it is less stable. But
metadata can be used to retrieve a stable identifier by querying a metadata
database.
And incdentally what you actually link to on IDEAL is an intermediate page which
has links to the abstract, PDF and the issue TOC. There are no external links
direct to the PDF.
Tony
"Dr. Guenther Eichhorn" <gei@head-cfa.harvard.edu> on 08/04/99 16:45:07
To: doyle@aps.org
cc: ref-links@doi.org, discuss-doi@doi.org, gei@head-cfa.harvard.edu (bcc:
Tony Hammond/AP/LDN/HARCOURT)
Subject: Re: [Ref-Links] Re: DOIs used for reference linking
Mark,
Yes, the ADS functions rbustly without DOI. As I mentioned,
the advantage of the bibcodes is that they can be user
generated, they are not opaque as DOIs. I don't think we
really need DOIs all that much, at least not in Astronomy.
We have two versions of links to the publishers, both are
algorithm generated, not handwritten. If we need to change
links to a publisher, we just need to modify the program that
generates the links.
One type of link to publishers assumes a name resolver at the
publishers site. This is implemented for instance for the
journals that are at the University of Chicago Press. They
have one address:
http://www.journals.uchicago.edu/cgi-bin/resolve
This address is called with the bibliographic code and the ID
for what part of the document is requested:
http://www.journals.uchicago.edu/cgi-bin/resolve?1980ApJ...241..915MPDF
This request will resolve to the link to the PDF file for the
article in the ApJ, volume 241, page 915.
The other type of link is more complicated, since we have to
build the address according to the publishers directory structure.
http://www.idealibrary.com/cgi-bin/links/citation/0019-1035/130/323
If the publisher decides to change the directory structure, we
would need to rebuild our link files. This is no big deal, since
these files are created through scripts. However, it would make
it easier if all publishers would provide name resolvers for
their on-line data. Whether these resolvers accept bibcodes,
DOI's, or other identifiers (or preferably both) doesn't really
matter for me, since I have to keep the concordance anyhow.
I have also implemented the capability of handling different
mirror sites of the publishers. For instance my link file
actually contains:
$AAS_JNLS$?1980ApJ...241..915M
for the above mentioned ApJ article.
When a user activates this link, my system replaces the publisher
identifier $AAS_JNLS$ with the user selectable mirror site of the
publisher. The user sets his/her preferences for mirror sites
once in the cookie driven preferences. From then on the selected
ApJ mirror sites is automatically linked to these references.
Since all this is handled through simple configuration files, it
is very easy to manage this for different mirror sites. Each ADS
mirror site has such a configuration file that sets the default
mirror sites for that ADS mirror. For instance the default ApJ
mirror at the US ADS site is the University of Chicago Press,
whereas the default ApJ site for our ADS mirror at the CDS in
France is the CDS mirror of the ApJ. The default handling is
completely transparent to the user. The user only has to set
things differently if for some reason s/he prefers to use a site
other than the provided default.
This has been very well accepted by our users as our usage
statistics show.
Guenther
PS: I am only subscribed to one of the lists, that's why I cross
posted.
---------------------------------------------------
Dr. Guenther Eichhorn | gei@cfa.harvard.edu
Project Scientist | Phone: 617-495-7260
Astrophysics Data System | Fax: 617-496-7577
Smithsonian Astrophysical Observatory
60 Garden Street, MS-83, Cambridge, MA 02138, USA
------ Original Message ------
In message <199904081504.LAA19715@next.aps.org>, Mark Doyle writes:
>Greetings,
>
>On Thu, 08 Apr 1999, "Dr. Guenther Eichhorn" wrote:
>
>> In the field of Astronomy we have had such a system for the last 5
>> years: The Astrophysics Data System (ADS). The ADS allows you to
>> identify articles by either Journal/Volume/Page or by searching for
>> authors or titles. The ADS then returns the information about the
>> article, together with a link to the on-line version at the publisher
>> (as well as other related links that are interesting for the end
>> users).
>[...]
>> Extending this system to opaque, publisher generated identifiers like
>> the DOI would be trivial. All that needs to be done is for the
>> publishers to include these identifiers in the data that they send to
>> the ADS. We would continue to maintain a concordance between the
>> publisher-generated opaque identifiers and the user-readable
>> bibcodes. This would make it very easy for users or other service
>> providers to automatically create links to specific articles in the
>> ADS and through that to the publisher.
>>
>> This system has been very successful in Astronomy and could easily be
>> extended into other fields.
>
>So what you are saying is that you have a stable, universal system that
>already exists that does not require DOIs to function robustly. What would
>adding DOIs achieve? The problem is that the rest of the STM world (well,
>the ST part anyway) doesn't have an ADS or Medline system. Hence the PubRef
>initiative.
>
>Do you store a URL for every article or dynamically generate them based on
>publisher URL templates and the metadata? If a publisher updates their
>scheme, do you have to update thousands for records or a few lines of code?
>
>Cheers,
>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
>
------------------------------------------------------
Ref-Links maillist - Ref-Links@doi.org
http://www.doi.org/mailman/listinfo/ref-links
------------------------------------------------------
Ref-Links maillist - Ref-Links@doi.org
http://www.doi.org/mailman/listinfo/ref-links