Re: [refbase-talk] Endnote/MODS XML output & attachments (was Re: (no subject))

[Chris P. <cdp...@uc...>]

On Wednesday 21 October 2009 1:57:57 pm Richard Karnesky wrote:
> > In the meantime, I contacted Chris Putnam who is the developer of
> > bibutils and he fixed the pdf handling  for bibutils.  From his point
> > of view, now the bibutils 4.4 can handle  pdf urls.
>
> CCing Chris on this.  Sorry for the rant, but 'internal-pdf://' was
> really a stupid invention.

Yes.  I simply treat "internal-pdf://" as a URL and let the user deal with
the fallout.  After the discussion below, perhaps this is suboptimal
(though I'll agree about the problem of a lack of a fully qualified URI...is
it possible to infer the position based on anything else in the EndNote
XML file, such as xxxx in <database path="xxxxx">?  I don't have 
access to EndNote to check.)

> Most programs require these URIs to be converted to fully qualified URI
> to allow you to import with file attachments.
> 
> Note also, that bibutils doesn't export this 'internal-pdf://' in the
> way that attachments are normally treated by other programs.  Indeed, I
> don't think there is explicit attachment support in bibutils yet.

Perhaps I'm really behind the times, but this Endnote issue was the first
time I had ever heard of "file attachments" in reference databases before.
 
> The scheme for MODS XML to support file attachments (supported by
> refbase & Zotero documented by the MODS standard) is to use an 'access'
> attribute in the URL branch, a'la:
>   <location>
>     <url displayLabel="Electronic full text" access="raw
> object">http://SERVER/PATH/TO/FILE.PDF</url>
>   </location>
> Bibutils does not currently say what kind of URL the link is to, so
> there is no way for others to know that it is an attachment.  (Of
> course, since the pseudo URI scheme is retained, no program other than
> EndNote could actually use the file even if the branch had the right
> attribute, but this may be beating a dead horse.)
> 
> One consequence of not having this explicit attachment support in the
> MODS part of bibutils, though, is that EndNote (and particularly those
> versions with poor or no EndNote XML support) may not be able to
> re-import the file attachments.  xml2irs will put the information in
> 'UR' instead of in 'L1'.
> 
> Explicit attachment support in bibutils would be valuable, particularly
> for use with programs that do use standard URI schemes.  This would
> need: the correct attribute for MODS input/output, above; using the 'L1'
> tag in RIS input/output, and using the
> "file = {Description:/file/path/or/uri/to.pdf:PDF}"
> scheme from BibTeX, etc.
> 
> > But  it will be helpful to know where and why the internal pdfs are
> > left out from refbase?
> 
> 'internal-pdf://' links are not handled because there is no sane way to
> do so.
> 
> 'L1' links in RIS are handled, though.  And explicit attachment support
> in bibutils may make importing attachments in other formats into refbase
> a bit easier.  Note, however, that there is no conversion of the URIs.
> There is also not presently a mechanism to upload file attachments at
> the same time you upload a file to import.  Having a good interface for
> this kind of thing poses a challenge & I'm not aware of any webapp that
> does this well.  If your file links already use the 'http://' scheme,
> they'll work.  But we haven't decided how we will support automatically
> adding attachments to the refbase files directory.  It is something a
> lot of people want, but we've yet to find ideas that are completely
> satisfactory for bulk import.
> 

This is quite helpful.  I was unaware that file attachments were being
handled by any other program other than Endnote.  Bibutils can start
handling "file"/"L1"/"Electronic full text" attributes in the formats now
that I know about them.  Certainly, one might even imagine screening 
URL's for ".pdf" extensions and "promoting" them to a "file attachment"
if that is observed.  I need to see how bibutils currently handles these
sorts of fields...hmm....it doesn't.  No surprise there.

Would bibutils conversion between these sorts of fields break refbase 
import in a bad way?

C.