Re: Asking for common wisdom on new field(s): References*
Le Sat, Nov 27, 2010 at 07:11:20AM +0900, Charles Plessy a écrit :
>
> In the case of bibliographic references, the needs goes beyond the use with
> than bibliographic software. For instance, the Blends web sentinel need only a
> part of the information in the BibTeX files, to provide a link to the upstream's
> publication.
>
> http://debian-med.alioth.debian.org/tasks/bio
>
> The question is then wether BibTeX files are the best medium for conveying
> fine-grained information such as the PubMed ID for instance. This is why
> I think that having the BibTeX as a central point of information is not
> convenient for that particular purpose. But I also understand that for
> the purpose of providing BibTeX files to the users, storing them as… BibTeX
> files in the source package is the most straightforward.
Dear Yaroslav,
thank you for your patience !
I have read the thread and had a week or two to think about it. I unfortunately
did not find time yet to push further the implementation of the
upstream-metadata system in the UDD (I tried to narrow down a bug in OpenOffice
instead, #601078). The code is ready in Subversion, but I just do not know the
procedure to get the UDD updated.
In the end, while using a common source of information, we are trying to do two
quite different things:
- On one hand, we want to provide a system-wide comprehensive
bibliography covering all the installed scientific software.
- On the other hand, we want to provide a short on-line reference when listing
the packages in our Blends tasks web sentinels, or on other sites like
packages.debian.org for instance.
I think that both tasks require a little bit of manual curation, which
means that maintaining two sources of information is not a big overhead.
- For the system-wide bibliography, we need to edit reference aliases to avoid
name collisions, and we need to remove the abstracts when they are non-free
material (see http://lists.debian.org/20071015014006.GB8951@kunpuu.plessy.org).
- For the on-line references, I think that we are limited by space on the
screen, and I am very tempted to limit the displaying to only one reference,
or alternatively an URL to upstram's list or instructions for citation when
the situation is more complex.
I think that I am currently entering more data than needed in the yaml file.
For online display, an authors list, a title, a journal name a year and a URL
should be enough.
The transfer of the bibliographic information out from the Blends task files is
one of my motivations for my proposition of using yaml files to feed the UDD. I
think that it is very important that sets of packages can be listed in single
Depends fields, in order to efficiently manage what we put in the tasks, and
that currently the bibliographic information is making that management
difficult. I would never have been able to quickly create the new med-cloud
task if package names would have been separated by half a dozen of lines of
metadata.
So here is what I propose:
- Since you have the motivation for collecting BibTeX information for a large
number of science packages, let's chose this format as a standard, and
document where the files should be in the source and binary packages.
- For the Blends task files, I will remove bibliographic information after it
is transferred to the UDD and the web sentinel generators are adapted to
this.
I would like to underline that this part is not undoing your current
blends-injection effort:
- It is currently useful as it has direct effect on our web sentinel pages.
- It will still be useful for the prospective packages, for which we do not
have other places to store metadata.
- I will use the injected bibliography as a the reference data for the
transfer to the UDD.
Have a nice day,
--
Charles Plessy
Debian Med packaging team,
http://www.debian.org/devel/debian-med
Tsurumi, Kanagawa, Japan
Reply to: