[Refdb-users] Re: nitpicking (element ordering in risx)

[Markus H. <mar...@mh...>]

Justus H. Piater writes:
 > Hi,
 > 
 > May I join into the discussion:
 > 

Sure, anytime.


 > Indeed, I am not sure what an input template would provide over and
 > above psgml-mode. Except: In my local copy of risx.dtd, I made the
 > citekey and the type #REQUIRED so that psgml-mode prompts for them.
 > 

This is not a bad idea for entering data from scratch. However, I
didn't want to force this upon users because refdbd can create useful
citekeys if none are provided. Some people may prefer this convenience.

 > Even better, one could convert the DTD to RELAX NG and use it with
 > nxml-mode which does XML validation.
 > 

I'm reluctant to go down this path as currently few tools support
RELAX NG. This is supposed to change eventually.

Before answering the detailed questions below, let me briefly mention
that the risx.dtd was not designed from scratch. Rather it is an XML
representation of the RIS tagged format with all its strengths and
(almost) all its weaknesses. The main purpose of this dtd is to have a
target for SGML/XML transformations. It was by no means written as a
replacement for any serious XML bibliography DTD like e.g. MODS.

 > Here are some questions and comments about risx.dtd:
 > 
 > - There is no editor element. How do I distinguish between authors and
 >   editors, say, for a book?
 > 

I've attempted to get some logic into the RIS author and title
levels. A book author is supposed to be encoded as AU/A1 whereas an
editor is an ED/A2. This does not translate well to the
part/publication/set distinction used in risx. Recent versions of risx
have a role attribute attached to the author element. However, this
may not yet be honored correctly during formatting bibliographies.

 > - I think there should be a way to provide a <url> for a conference
 >   article that's been published in a proceedings volume. This belongs
 >   into the <part>, I think.
 > 

Question to Bruce (he knows MODS better than me): can MODS do this?

 > - This may have been answered elsewhere, but I can't remember right
 >   now: How are full vs. abbreviated journal titles handled by RefDB?
 > 

Journal titles are kept in a separate table. All references encoding
articles from this journal can use either the full or abbreviated name
depending on what the bibliography style requires. There are fallbacks
if one of the titles is missing.

Currently there is no convenience command to maintain the journal
title list. If you have references containing only one type of title,
it is best to retrieve them, add the missing title, and update the
references.

 > - It would be nice to provide an event address for conferences (which
 >   is usually different from the publisher's or organization's
 >   address).
 > 

Bruce: can MODS do this?

 > - I think we need a mechanism to protect individual characters against
 >   automatic case conversion of titles by style sheets (like {G}aussian
 >   and {HMM} in BibTeX).

This would only be a problem if there are styles that force all
lowercase. citestylex.dtd supports this but I don't know whether any
real-life journal requires this. The other possibilities (all caps or
keep as is) would work ok if you supply the titles in proper mixed
case.

If this is not sufficient we'll have a problem. 

 > 
 > - In <libinfo>, why isn't <reprint> optional? Also, the only allowed
 >   content is <date>, not even #CDATA is permitted. This contradicts
 >   the RIS specification of the RP field, doesn't it?

I think you run a reference manager mainly to keep track of your
offprints/electronic copies. The reprint status is kind of essential
to do this. But if this bothers people I won't have a problem changing
this element to optional.

BTW the RIS spec says the RP field can contain one of three status
notes, one of which (ON REQUEST) may be followed by a date. This is
exactly what the reprint element represents: The attribute encodes the
status, the optional child is the date.

 > 
 > - Is there really no established standard for representing references
 >   in XML? By TEI, for example, or something linked to the Dublin
 >   Core... And what is the relation of RIS/RISX/RefDB to Bibliofile?
 > 

The nice thing about standards is that there are so many to choose
from (don't know whom to attribute this to). I'm still evaluating
whether MODS is suitable for the purposes of a reference
manager/bibliography tool. It was clearly not designed for this
purpose but it probably gets closest. 

 > Another issue: I'd like the "master" representation of my references
 > to be an XML file rather than a data base because that allows me to
 > choose and change the format as desired. To make this practical, it
 > would really help if RefDB could optionally identify references by the
 > citekey instead of by the numeric ID. This would allow me to store
 > references from different sources in the same RefDB database, and to
 > update some of them by re-importing manually maintained XML files.
 > 

I don't know what you mean by "choosing and changing the format"? Do
you want to move your data back and forth between different DTDs?
Otherwise XML is plain text that you can (with a few exceptions)
format any way you want, even if it originates from a database. In any
case I don't understand the advantage of maintaining your references
outside of a database. You can retrieve them from the database as XML
files and update them any time. You can even retrieve the full
database as XML periodically and check this into CVS. But the point of
a reference manager is to make the references accessible by simple
queries. If you maintain your data outside the database, you're back
to grep, or you overload your brain with things a database was
designed to remember.

The ID and citekey are mostly interchangeable. Both must be unique in
a database. The ID is automatically created by the database engine and
as such comes for free. But you can of course update references by
providing only the citekey but no ID. The risx import routine always
checks the citekey first, then the ID.

regards,
Markus

-- 
Markus Hoenicka
mar...@ca...
(Spam-protected email: replace the quadrupeds with "mhoenicka")
http://www.mhoenicka.de