On Nov 25, 2005, at 3:12 PM, David Wilson wrote:
In any of these groups the user should be able to select the sort
order name,
date / date, name etc.
Hmm ... I'm finding myself really not liking this idea. It adds
significant complexity all around (XML, processing code, GUI), for what
benefit? Why would someone want or need to have different sorting
within sections of a reference list?
And as Martha suggested there could be an arbitrary
number and levels of headings, and sorting options. (Published /
Un-published
could be sorted based on record content tests ie ' If Publisher Name
exists
then work is a member of 'Published'. If NOT 'Published' then work is
a
member of 'Unpublished'.)
If you want to make this argument, then I think you're going to need to
back it up with a lot of research. I'm not convinced. And I don't
think we should support every style out there. We should focus on
fully supporting the dominant styles: APA, Chicago, MLA, Blue Book,
etc.
Also I have suggested before we need a mechanism to allow to manually
over-ride the sorting order ...
Example?
For setting up the bibliography table I suggest we have a GUI that
allows the
user to set Heading Groups and table sorting options and “pre-sorting
rules”.
For pre-sorting to operate each bibliographic record would have a
name-sort
field which the sort-pre-processing would convert the name to the
sort-name.
Am not understanding you here David. Can you explain again?
An example of special sorting rules for names: how to sort ‘Mujahid
Usamah Bin
Ladin’ ; Bin Ladin is the family name but ‘Bin’ should always ignored
in
sorting arabic names, and the name would be grouped in the ‘L’s.)” or
the
generally known all 'Mc' or 'Mc ' are replaced with 'Mac'. Also you
may want
to Anglicise the names by replace all the umlauts or accents with the
standard unaccented character. There would be a standard set of
pre-sort
rules but the user could add extra rules. These could be rules to fix
Chinese
– English transliteration problems ie change all 'Mao Tse Tung' to 'Mao
Zedung'. To fix the situation where the rules do not seem to work,
force the
sorting order - For record ID 34334453 sort-name ='smith, d. h."
OK, I see what you're saying. I don't think this has anything to do
with either CSL or the GUI. It needs to be handled in the intersection
of the metadata and the formatting engine. E.g., you encode a name
with a language attribute, and the engine needs to know how do deal
with language.
There's a reason I leave this loose in CSL; if I start allowing this
fine-grained configuration, everything gets way too complex.
So how would this interact with the CiteProc formatting engine ? The
Bibliographic Table GUI would send a list of citation ID's to
CiteProc, and
CiteProc would return the formatted citation strings to the selected
style.
The sort order does not matter at this point. Through the GUI process
the
pre-sort rules populate the sort-name fields and the Headings are
defined.
When the user has finished the bib table setup and pressed the OK
button to
generate the table. The GUI process could return to Citeproc the
sorted list
of citation Ids along with the Headings. ie
Bib Heading level1='Primary Sources', CiteIds=1234, 13445, 234234,
234234,
234234
Bib Heading level2='Published Sources' CiteIds=
45234,23423,2344,3566,576567,
Bib Heading level2='Unpublished Sources' CiteIds=
576567,56758,3245,123,234,4223,8645
Bib Heading level1='Secondary Sources' CiteIds= 463456923, 238492,
2348974,
088776
This should be easier to work with.
I don't see that. To me this is needlessly complex. I've never really
used OOo' bib feature. My frame of reference is Endnote + Word and
BibTeX + LaTeX, and in both cases, reference list generation is fully
automatic. In Endnote, you can set it to dynamically add references as
you add citations. It doesn't have grouping, but that seems to me a
separate issue.
Regarding Heading groups. I think that user defined groups can be just
assigned to the cited works in a document. If a user then assigns both
'Primary Sources' and 'Secondary Source' to work then it will appear
in both
lists.
But again, you're adding complexity. If you allow one group, then that
might suggest a single pop-up field that allows a user to associated an
item with a group. If you allow arbitrary, then you need to allow
arbitrary pop-ups in the GUI.
In general, every single attribute parameter or element I add means
more complexity. We need to be careful before doing that.
Example, right now I don't allow configuring sort order in different
reference classes, except with simple parameters. Why? Because the
only place where multi-level sorting matters is in the author-year
class. And the author-year class by definition sorts by author, year,
month-day. So why make life more complicated by allowing -- and indeed
requiring -- people to manually set this?
Grouping isn't even supported in existing commercial competition. And I
don't think Endnote can even do some of the things I have in CSL and
CiteProc.
Now, you might be wondering how I imagine the interaction of the config
process and formatting.
Reference list generation should be similar to how Endnote works; it
should be fully automatic and dynamic by default. If I add a citation,
it ends up in the reference list in almost-realtime.
Grouping? Defined by a separate panel in the config, where one might
have:
group label
----------------------
primary Primary Sources
secondary Secondary Sources
As the above suggests, I have in mind a pretty simple GUI. If we add
the ability for arbitrary depth, that GUI gets a lot more complicated.
What do other people think?
Bruce
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
