%% @latexfile{
%% filename="rhead.ltx",
%% version="1.0",
%% date="22-Dec-1993",
%% filetype="TUGboat article",
%% copyright="Copyright (C) TeX Users Group.
%% Copying of this file is authorized only if you make
%% absolutely no changes to your copy. Copies may be
%% distributed without charge provided this notice is
%% preserved on all copies.",
%% author="David Rhead",
%% address="Cripps Computing Centre,
%% University of Nottingham,
%% Nottingham NG7 2RD,
%% United~Kingdom",
%% email="Internet: d.rhead@vme.nott.ac.uk",
%% codetable="ISO/ASCII",
%% checksumtype="line count",
%% checksum="1024",
%% keywords="bibliographic references, LaTeX",
%% abstract="This file contains the revised text of the article
%% `A suggested ``operational requirement'' for \LaTeX3's
%% treatment of bibliographic references', which appeared
%% in TUGboat 14(4), December 1993, pp.425-433, with an
%% update in TUGboat 15(1), March 1994, pp.xx-xx.
%% References for this article, in the form of a BibTeX
%% file, are contained in the file rhead.bib.
%% Comments should be directed to the author, not to
%% TUGboat or to the TeX Users Group.",
%% notes="This file should be deleted from the archive upon
%% the general release of LaTeX3."
%% }
%%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% rhead.ltx
\documentstyle{ltugboat}
\vol 14, 4. % volume, issue.
\issdate December 1993. % month, year of publication
\PrelimDraftfalse
\newcommand{\bs}{\char`\\}
\title{A suggested ``operational requirement'' for \LaTeX3's
treatment of bibliographic references%
\thanks{\copyright\ \TeX{} Users Group\hfil\break
This paper is a revised version of an article that appeared in
\tubissue{14}{4}, December 1993, pp.~425--433, with an update in
\tubissue{15}{1}, March 1994.}}
\author{David Rhead}
\address{Cripps Computing Centre
University of Nottingham
Nottingham NG7 2RD
United~Kingdom}
\netaddress{d.rhead@vme.nott.ac.uk}
\begin{document}
\maketitle
\begin{abstract}
It is suggested that:
\begin{itemize}
\item \LaTeX3 should aim to support the principal citation schemes used
in conventional publishing
\item consideration be given to a {\it modus vivendi\/} between \LaTeX3 and
mainstream bibliography-formatting software.
\end{itemize}
\end{abstract}
\section{Introduction}
Ideally, when writing software, it is a good idea to write down what the
software is intended to achieve\Ldash the ``operational requirement''\Rdash
before writing any code. (In the remainder of the article, ``operational
requirement'' is abbreviated to ``OR''.)
This article attempts to take an OR approach to the ``bibliographic
reference'' aspects of \LaTeX3.%
\footnote{Obviously, there are limits to the what the OR approach can
achieve. For example, it is difficult to quantify ``usability''.
Nevertheless, the approach should facilitate debate about objectives {\em
before\/} the ``user interface'' has been fixed.}
The objective is to stimulate debate\Dash if you don't agree with my
suggestions, please suggest specific alternatives!
Generalizing the approach taken by the \LaTeX\ 2.09 manual \cite[pp.\
73--74]{lamport-86}, it is convenient to divide the topic into ``doing it
yourself'' and ``using bibliography-formatting software''.
\section{Doing it yourself}
\label{DIY}
In effect, the only scheme that is ``fully supported'' by \LaTeX\ 2.09 is
``reference by number, where the sequence of numbers is determined by
position in the reference-list''.
By contrast, for ``real-world publishing'', my impression is that:
\begin{enumerate}
\item
only a minority of ``instructions to authors'' specify anything like
the default \LaTeX\ 2.09 scheme. This minority consists of those
journals that specify ``reference by number, with the reference-list
in alphabetical order of author's names''.
\item the majority of ``instructions to authors'', style-books, etc.,
specify one of the following:
\begin{enumerate}
\item
reference by number, with the reference-list in order of first
citation
\item
author-date
\item
``short-form in footnotes''. For publications in the humanities,
there seem to be two main variants of this scheme, depending on
whether or not there is a reference-list.%
\footnote{If there is no reference-list, the convention is usually ``first
citation gives full bibliographic details, subsequent citations give
cross-reference to first citation''. This variant is common in law
publications, when it is used in conjunction with numerous law-specific
citation conventions \cite{harvard-law}.}
\end{enumerate}
{\SMC ISO} 690 \cite[sec.\ 9]{iso-690} provides a convenient specification
of the details of these schemes. The default \LaTeX\ 2.09 system gives
no particular help to anyone wanting to use them.%
\footnote{%
\BibTeX\ can help with (a).
Anyone wishing to use (b) will probably grope around in archives
looking for style-options that: arrange for {\tt \bs cite} to give (\dots)
rather than [\dots]; omit [\dots] from the reference-list; support
date-only citations when the author's name appears naturally in a
sentence.
Apart from the law-specific Lexi\TeX\ \cite{bennett-93}, I'm not aware of
any 2.09-related software that helps people who wish to use scheme (c).}
\item a few publishers specify alternative schemes.
E.g.,
\begin{itemize}
\item some Springer journals%
\footnote{See the ``instructions for authors'' in, for example, {\it
Mathematische Zeitschrift.}}
accept citations of the form ``first letter of author's surname,
in square brackets''
\item
Butcher's {\it Copy-editing\/} \cite{butcher-92} mentions a variation
of the reference-by-number system in which there is a separate
numerical sequence for each letter, and a variation of the
author-date system in which a number is used instead of a date
\item a scheme like the \BibTeX\ {\tt alpha} style is sometimes
used (for example, in the journal {\it Formal Aspects of
Computing\/}).
\end{itemize}
\end{enumerate}
\noindent
Therefore, I suggest that the OR for \LaTeX3:
\begin{itemize}
\item
should include support%
\footnote{I assume that ``sorting a reference-list'' will be beyond the
scope of \LaTeX3. Thus, in practice, the \LaTeX3 ``support'' might be
minimal (a ``better than nothing'' warning that a reference-list needs
human intervention, perhaps). People who want anything better would
be advised to use bibliography-formatting software.}
for the schemes mentioned in items 1 and 2 above, i.e.,
\begin{itemize}
\item
a 2.09-like scheme aimed at journals that specify ``reference by
number, with the reference-list in alphabetical order of author's
names''
\item
the schemes specified in {\SMC ISO} 690, namely: ``reference by
number, in order of first citation'', author-date, and 2 variations
of ``short-form in footnotes''.%
\footnote{To support these schemes, it is probably desirable that \LaTeX3
should be able to determine whether a citation of a source is ``the first
citation'' of that source. Clearly this would help to provide support for
``reference by number in order of first citation''. In the author-date
case, it would allow support for the convention \cite[sec.\ 3.87]{apa-83}
that, when there are multiple authors, they should all be named in the
first citation but ``{\it et al.}''\ should be used subsequently. It
might also help to provide support for the variant of the short-form scheme
in which a ``subsequent citation'' uses the short-form and gives a
cross-reference to the footnote containing the ``first citation'' (where
full details of the source can be found).}
\end{itemize}
\item
should bear in mind the possibility of a ``plug-in module'' to support
law conventions. Since such conventions are crucial only to lawyers,
it would probably be inappropriate to delay \LaTeX3 while
law-specific commands were finalised, or to increase the bulk of the
\LaTeX3 manual by including law-specific material. Nevertheless, it
might be worth simultaneous experiments with a prototype \LaTeX3
and a prototype law-support module, in the hope that the law-specific
commands in such a module might end up with a similar ``look and
feel'' to those for the mainstream ``short-form in footnotes''
commands.
\item
need not include support for the alternative schemes mentioned in
item~3 above (although the possibility of ``plug-in modules'' to
support these schemes might be borne in mind).
\end{itemize}
\noindent
In addition, the following features are desirable:
\begin{itemize}
\item
for situations where several sources are cited simultaneously:
\begin{itemize}
\item
a syntax that permits a particular division of each source to be
pin-pointed \cite[sec.\ 15.25]{chicago-82}. (The \LaTeX\ 2.09
\verb+\cite[...]{...}+ syntax only supports pin-pointing within a
single source.)
\item
some mechanism for sorting reference-by-number citations into
ascending numerical order \cite[p.~106]{acs-86}.%
\footnote{If it is not feasible to sort reference-by-number and author-date
citations into a desired order within \LaTeX3, some mechanism for providing
warnings if simultaneous citations are in the wrong order would be ``better
than nothing''.}
\item
some mechanism for sorting author-date citations%
\addtocounter{footnote}{-1}\footnotemark\
into alphabetical
order of author's surnames (or, ideally, the order in which the
sources appear in the reference-list) \cite[sec.\ 3.91]{apa-83}
or into ``date of publication'' order \cite[sec.\
15.24]{chicago-82}.
\end{itemize}
\item
support for types of bibliography that, although not as common as a
single undivided list, are appropriate in particular circumstances,
namely:
\begin{itemize}
\item a list divided into sections according to kinds of material,
subject matter or other appropriate categories
\item an annotated bibliography
\item a bibliographical essay.
\end{itemize}
See, for example, the {\it Chicago Manual of Style\/} \cite[chap.\
15]{chicago-82}.
(\LaTeX\ 2.09's {\tt thebibliography} environment has a name which
suggests that it is the natural choice for {\em any\/} bibliography.
Unfortunately, its structure isn't really appropriate for everything
that might reasonably be called ``a bibliography''.)
\end{itemize}
The above might provide the major elements of an OR\@. Minor elements may
be more difficult to specify, but can perhaps be summarized as
\begin{quote}
\LaTeX3 should be able to survive $\beta$-testing of whether
it can conveniently deliver bibliographic details formatted as
specified by influential style-books and ``instructions for authors''.
\end{quote}
See
\cite{harvard-law,iso-690,butcher-92,apa-83,chicago-82,mhra-91,mla-88,%
vancouver-91,bs-5605,bs-1629,bs-6371}.
\subsection*{Multi-author documents}
I think it desirable that \LaTeX3's successors to the ``standard styles''
should support multi-author documents (e.g., a journal-issue made up of a
number of articles, or a conference-proceedings made up of several
contributions).
Hence:
\begin{itemize}
\item it should be possible to have several reference-lists within a single
document
\item
there should be allowance for the possibility that a single document
may use two or more citation schemes. E.g., since the ``instructions
for authors'' in {\it Mathematische Zeitschrift\/} gives a choice of
three citation schemes, an issue of the journal may involve three
distinct schemes.
\end{itemize}
\section{Using bibliography-formatting software}
\subsection[Background]{Background\footnotemark}
\footnotetext{Warning: I do not currently have ``hands on'' experience of
using \LaTeX\ in conjunction with software other than \BibTeX\ (although I
have browsed through as many of the relevant manuals as I could find).
Hence, the ideas given in this section, and in section \ref{misc},
are theoretical and speculative.}
\label{background}
\subsubsection{Software available}
The bibliography-formatting software that is ``advertised'' in the \LaTeX\
2.09 manual is \BibTeX\ \cite{lamport-86,patashnik-88}. Tib
\cite{alexander-89} is also sometimes mentioned in \TeX\ circles.
In fact, there are a large number of programs
available for generating and formatting reference-lists.
A recent review article \cite{stigleman-93} names 52 such
programs.
Judging by comments on the {\tt bibsoft} list, the most important
bibliographic programs (from the point-of-view of professional librarians
and bibliographers) seem to be EndNote, Library Master, Papyrus, ProCite
and Reference Manager. (Appendices \ref{suppliers} and
\ref{discussion-lists} give details of the {\tt bibsoft} list and of the
relevant vendors.)
Of these, EndNote, Papyrus, ProCite and Reference Manager have procedures
for processing a ``manuscript'', filling in the in-text citations, and
generating the corresponding reference-list. Although I understand that a
similar facility is planned for the next version of Library Master, I don't
know what form this will take. Therefore, when referring to these
programs, I will use:
\begin{description}
\item[``main~4''] to mean the programs (EndNote, Papyrus, ProCite and Reference
Manager) whose procedures for filling in the in-text citations are
currently known
\item[``main~5''] to mean the ``main~4'' plus Library Master.
\end{description}
From a \LaTeX-er's point-of-view, the public-domain \BibTeX\ and Tib are
obviously attractive, since they were {\em designed\/} to work with
\TeX/\LaTeX, and are available for most of the platforms on which
\TeX/\LaTeX\ are available. By contrast, the ``main~5'' are:
\begin{itemize}
\item proprietary
\item currently aimed at ``wordprocessor'' users%
\footnote{Certain vendors state that \TeX\ is one of their program's
``supported wordprocessors''. You may or may not regard this as a hopeful
sign!}
\item only available on a restricted selection of platforms. (All are
available for {\SMC MS-DOS}\@. Some are available for Macintosh or
{\SMC \VAX/VMS}.)
\end{itemize}
Nevertheless, there are many things about the ``main~5'' that are of
interest:
\begin{itemize}
\item
The programs have standard procedures for importing information from
online information services, {\SMC CD-ROM}s,
library catalogues and popular database programs.
\item
They generally have good facilities for maintenance of a ``personal
bibliographic database'', and for searching such a database for
entries that satisfy particular criteria.
\item
It seems likely that the programs will continue to be developed and
supported into the future. (By contrast, my understanding is that
\BibTeX\ will be ``frozen'' when version 1.0 has been finished.)
\item
There is a choice. If one program has underlying assumptions that
don't match the assumptions that are usual in your discipline, you can
look for an alternative!
\end{itemize}
Even if you don't regard the ``main~5'' as of positive interest, you may be
unable to avoid them. If a research group contains a \LaTeX-ing minority
and a non-\LaTeX-ing majority:
\begin{itemize}
\item
the group's
``standard bibliography-formatting software''
may {\it de facto\/} be one of the ``main~5'',
because that is what ``the majority'' use
\item
the \LaTeX-ers will then be at a serious disadvantage if they cannot
use their colleagues' bibliographic databases.
\end{itemize}
Also, if your librarian is providing bibliographic information in
electronic form (e.g., from a computerized library catalogue), s/he may
offer an off-the-shelf way to get the information into a database for one
of the ``main~5'', but be unable to help you if you use \BibTeX.
Overall, it seems to me desirable that, as well as having standard
procedures for inter-working with \BibTeX\ and Tib, \LaTeX3 should have
standard procedures for inter-working with the ``main~5''. Such procedures
are unlikely to be perfect, but it should be possible to agree on some {\it
modus vivendi}.%
\footnote{It is unlikely that the vendors will re-focus their products to
concentrate on \LaTeX\ users\Dash and equally unlikely that \LaTeX-ers will
start to think of themselves as ``wordprocessor users''. Nevertheless,
with a few minor changes (which might involve the \LaTeX\ end, the
bibliographic program end and/or the documentation), it should be possible
for \LaTeX3 and the mainstream bibliographic software to work reasonably
well together.
``{\it Modus vivendi\/}'', i.e., ``an arrangement between peoples who agree
to differ'', seems to fit the situation quite well.}
\subsubsection{{\it Modus vivendi\/} with the main~4?}
\label{modus-vivendi}
Before considering how \LaTeX\ might co-operate with the ``main~4''%
\footnote{Hopefully, it will be possible to use the same general ideas for
Library Master when its procedure for ``filling in the in-text citations
and generating the reference-list'' becomes known.}
it is convenient to contrast \BibTeX's approach with that of Tib.
\BibTeX's approach involves searching a \LaTeX\ {\tt .aux} file for details
of in-text citations, and then writing out a {\tt .bbl} file. The {\tt
.bbl} file defines a reference-list that is read in when \LaTeX\ is next
applied to the root file.
Tib's approach is different. It starts with a {\tt .tex} file that
contains ``incomplete or keyed citations'' within citation-delimiters, and
produces another {\tt .tex} file that contains proper in-text citations
plus (optionally) a reference-list.
When the procedures used by the ``main~4'' are interpreted in terms of
\LaTeX, they seem to be more akin to Tib's approach than to \BibTeX's. It
looks as though the end-user would start with a {\tt .tex} file containing
keys, etc., within citation-delimiters, and use the bibliography-formatting
program to produce a near-duplicate {\tt .tex} file that contains proper
in-text citations plus a reference-list.
In fact, Tib's citation-delimiters are chosen so that:
\begin{quote}
The escape characters of Tib do not interfere with \TeX\ processing. If
\TeX\ is applied to the original pre-Tib document, the escape characters
and incomplete citations will appear as written.
\end{quote}
I.e., the pre-Tib {\tt .tex} file and the post-Tib {\tt .tex} file are both
valid \LaTeX\ input files.
This seems a useful precedent. If \LaTeX\ could inter-work with the
``main~4'' in an analogous way, it would not be necessary to
\begin{quote}
Apply bibliography-formatting software.\\
Then apply \LaTeX.
\end{quote}
every time that a {\tt .dvi} file is required. For example, if someone is
concentrating on getting their equations typeset correctly, they might want
to get {\tt .dvi} files quickly without always having to go through the
bibliography-formatting step. At the equation-checking stage, they may
just want a {\tt .dvi} file that shows their equations, and not be worried
about the appearance of their in-text citations or reference-list.
A potential problem for any \LaTeX-er trying to follow the Tib precedent,
is that EndNote and ProCite use \verb+#+ to identify ``number within
database''. Hence the end-user may need to put a \verb+#+ (which is one of
\LaTeX's 10 ``special characters'') within the relevant
citation-delimiters. (See Table \ref{analogues} for details of the
programs' default citation-delimiters, and the alternatives available.)
\begin{table*}
\begin{center}
\begin{footnotesize}
\begin{tabular}{lll}
\hline\hline
Software & Citation & Notes \\
& delimiters & \\
\hline\hline
\LaTeX\ 2.09 with \BibTeX & {\tt \bs cite\{\ \}} & \\
\hline
Tib & {\tt [.\ .]} &
The delimiters {\tt <.\ .>} are used \\
& & in some circumstances \\
\hline
EndNote & Default: {\tt [\ ]} &
You can tell EndNote to look for alternative \\
& & 1-character delimiters (e.g., {\tt <\ >}) \\
\hline
Library Master & Not known &
I understand that a facility for ``given \\
& & the in-text citations, compile a \\
& & reference-list'' is in preparation \\
\hline
Papyrus & Default: {\tt \%\%\ \%\%} &
You can tell Papyrus to look for alternative \\
& & delimiters (but ``start delimiter'' must \\
& & be the same as ``end delimiter'') \\
\hline
ProCite & Default: {\tt (\ ) } &
You can tell ProCite to look \\
& & for {\tt [\ ]} rather than for {\tt (\ )} \\
\hline
Reference Manager & Default: {\tt \{\ \}} &
You can tell Reference Manager to look \\
& & for alternative delimiters. ``Start delimiter'' \\
& & and ``end delimiter'' can each have up \\
& & to 7 characters. \\
\hline\hline
\end{tabular}
\end{footnotesize}
\end{center}
\caption{Citation-delimiters: defaults and alternatives\label{analogues}}
\end{table*}
One way of imitating Tib (in spite of the possibility of \verb+#+
characters) might be to arrange delimiters such that the proprietary
program's ``start delimiter'' is interpreted by \LaTeX\ as being equivalent
to \LaTeX\ 2.09's \verb!\verb+!, and its ``end delimiter'' is interpreted
as equivalent to the \verb!+! that terminates the text introduced by
\verb!\verb+!. Then:
\begin{itemize}
\item
if \LaTeX\ is applied to the original {\tt .tex} file, the citation
keys will be typeset ``as is'' in a {\tt typewriter} font (to remind
the \LaTeX-er that the bibliographic software needs applying before
the document can be regarded as finished)
\item
if the bibliographic software is applied to the original {\tt .tex}
file, a new {\tt .tex} file will be produced that, when \LaTeX-ed, has
proper in-text citations and a reference-list.
\end{itemize}
Overall, the \LaTeX-er will be able to apply \LaTeX\ and the bibliographic
software in either order (in much the same way that \LaTeX\ and Tib can be
applied in either order).
This approach could be the major element of a {\it modus vivendi\/} between
\LaTeX3 and the ``main~4''. Table \ref{choice-of-delimiter} shows some
delimiters that might be suitable.
\begin{table*}
\begin{center}
\begin{footnotesize}
\begin{tabular}{llll}
\hline\hline
Biblio.\ & Tell bib.\ software & Tell \LaTeX3 & Notes \\
software \\
\hline\hline
EndNote & Delimiters are & {\tt<}\dots{\tt>} is equivalent \\
& {\tt <} and {\tt >} & to 2.09's {\tt \bs verb+}\dots{\tt+} \\
\hline
Papyrus & Delimiter is {\tt"} & {\tt"}\dots{\tt"} is equivalent \\
& & to 2.09's {\tt \bs verb+}\dots{\tt+} \\
\hline
ProCite & & & No obvious alternative to ``always \\
& & & apply ProCite before \LaTeX'' \\
\hline
Reference & Delimiters are &
{\tt \bs bsoft\{\dots\}} is equiv.\ to \\
Manager & {\tt \bs bsoft\{} and {\tt\}} &
2.09's {\tt \bs verb+}\dots{\tt+} \\
\hline\hline
\end{tabular} \\[2mm]
\begin{tabular}{lp{100mm}}
{\it Note:}
& Clearly the default Papyrus and Reference Manager delimiters (see
Table \ref{analogues}) must be changed if the end-user is to have the
option of applying \LaTeX\ without having previously dealt with
citations, etc. However, the Papyrus and Reference Manager keys are
not liable to contain a {\tt \#} character. Hence, it is not crucial
whether Papyrus and Reference Manager keys are ``hidden'' from \LaTeX.
\end{tabular}
\end{footnotesize}
\end{center}
\caption{Choice of delimiters for {\it modus vivendi\/}?%
\label{choice-of-delimiter}}
\end{table*}
A {\it modus vivendi\/} would also need to incorporate an approach to the
``root file and {\tt \bs include}-ed files'' situation. Although I don't
have any specific suggestions at this stage, I speculate that support for
this feature might be obtained by reference to the bibliographic software's
support for analogous features in wordprocessors (e.g., WordPerfect's
``master document and subdocument'' scheme, and Microsoft Word's
``include'' scheme).
\subsubsection{Preferred interface}
The suggestions in Table \ref{choice-of-delimiter} are intended as part of
a {\it modus vivendi\/} between \LaTeX3 and the {\em current\/} versions of
the ``main~4''. Although the general approach is the same, the details
differ from product to product.
It would be open to \LaTeX-ers to decide on a preferred interface, and to
inform the vendors of their preference in the hope that it may be possible
to implement the approach more consistently at some time in the future. We
wouldn't lose anything by asking!
For example, if the preferred interface involved \verb+\bsoft{+{\it
key\/}\verb+}+ (as shown in Table \ref{choice-of-delimiter} for Reference
Manager), it would be open to us to ask the other vendors to relax their
rules on citation-delimiters so that future versions of the ``main~5'' will
all accept \verb+\bsoft{+{\it key\/}\verb+}+. If we were lucky enough to
get the vendors' agreement, we might be able to produce notes about ``using
proprietary bibliographic software with \LaTeX'' that would appear simpler
to the end-user than Table \ref{choice-of-delimiter}.
\paragraph{Note}
It might actually be possible to have a {\it modus vivendi\/} (e.g., with
Reference Manager) that involves \verb!\verb+!{\it key\/}\verb!+!, rather
than having an additional command such as \verb+\bsoft+ (which would, in
any case, be implemented in much the same way as \verb!\verb!). The
bibliographic software will probably ignore things within \verb!\verb+!
and \verb!+! that don't look like citation keys. Nevertheless, I would be
inclined to introduce an extra command (e.g., \verb+\bsoft+) so that {\tt
.tex} files can be ``marked up logically'' to distinguish between:
\begin{itemize}
\item delimiters for a key that is intended for processing by
bibliographic software
\item delimiters for text that is intended to appear in a {\tt typewriter}
font in the final document.
\end{itemize}
\subsubsection{Hybrid approaches 1}
One can envisage schemes that embed a proprietary bibliographic system's
mechanism for dealing with citations and reference-lists within \LaTeX's
mechanism (or {\it vice versa\/}). Examples might include:
\begin{itemize}
\item
changing the Papyrus delimiter to \verb+!!+, and then
putting Papyrus delimiters inside a \LaTeX\ \verb+\cite+ command,
thus \verb+\cite{!!+\dots\verb+!!}+.%
\footnote{Bernard J. Treves Brown, of Manchester University, is
experimenting with this technique.}
\item
trying to get proprietary bibliographic software to read an {\tt .aux}
file, and write a {\tt .bbl} file, as \BibTeX\ does. (Perhaps this
could be done by a shell script which invokes the proprietary software
in a suitable way.)
\end{itemize}
Generally, I fear that such hybrid schemes may lead to confusion, and I
would not be inclined to pursue them:
\begin{itemize}
\item
Anyone constructing a hybrid scheme will have to be very careful about
``which software is in charge when'' (e.g., whether citation numbers
are incremented by \LaTeX, by the proprietary system, or by ``one
shadowing the other''). The hybrid scheme will need maintenance
(e.g., someone will need to verify that the scheme still works with
each new release of the proprietary system). There may be three lots
of documentation for the end-user to study: that about \LaTeX3, that
about the proprietary system, and that about the hybrid scheme's
subtle combination of elements of both. If anything goes wrong, it
may be in ``a grey area'', which is neither the responsibility of the
\LaTeX3 project, nor the responsibility of the bibliographic
software vendor.
\item
The proprietary systems seem more akin to Tib than to \BibTeX. To try
and force them into the \BibTeX\ stereotype when they are not designed
to work like \BibTeX\ seems like ``asking for trouble''. I doubt
whether the \TeX\ community has the resources to produce interfaces
that ``make proprietary systems work like \BibTeX'', and I doubt
whether the vendors have the inclination to commit such resources.
\end{itemize}
My instinct is that it would be better to have a simple interface (e.g.,
conventions such as those outlined in Table \ref{choice-of-delimiter}), so
as to put the end-user in a situation where responsibilities are clear:
\begin{itemize}
\item
typesetting is the responsibility of \LaTeX3
\item
bibliography-generation is the responsibility of the bibliographic
software.
\end{itemize}
Hence, if using a proprietary bibliographic system, the end-user should
ignore the \LaTeX3 manual's descriptions of commands for ``doing it
yourself'' (i.e, ignore the \LaTeX\ commands envisaged in section
\ref{DIY}), and ignore anything that is provided to support people
using \BibTeX:
\begin{itemize}
\item
The proprietary system will be ``in charge'' of bibliography
generation. The method used will be that envisaged by the vendor, and
documented in the vendor's manual: if it's good, the vendor will get
the credit; if it's bad, the vendor will get the blame.
\item
The delimiters in the {\tt .tex} file will be delimiters for the
proprietary system (chosen, if possible, in such a way that the {\tt
.tex} file is acceptable to \LaTeX\ even before processing by the
proprietary system). They might be as shown in Table
\ref{choice-of-delimiter}. The ``keys'', etc., inside the delimiters
will follow the rules given in the vendor's manual ({\em not\/} the
rules given in the \LaTeX3 manual about keys when ``doing it yourself'').
\item
The proprietary system will be ``told to produce \TeX\ output''.
How good or bad it is at this will be the responsibility of the
proprietary system (although interested \LaTeX-ers might advise the
vendors about what is required).
\end{itemize}
Overall, the end-user will get in-text citations filled in, and
reference-lists generated, in the standard way that is described in the
manual that describes the proprietary system. If this standard way does
not suit a \LaTeX-er's requirements, it may be better for him/her to seek
alternative bibliography-formatting software rather than spending time
trying to circumvent the problems.
Of course, if people want to put effort into developing hybrid schemes, and
happen to get good {\it modus vivendi\/} between \LaTeX\ and proprietary
bibliographic systems, I would be delighted to find that my instinct is
wrong!
\subsubsection{Hybrid approaches 2}
Other hybrid approaches might use a proprietary system and \BibTeX\ ``in
series'':
\begin{itemize}
\item
One might regard the proprietary system's database as a ``staging
post'', where information stays briefly before being converted to a
\BibTeX\ database. For example, if a proprietary system can import
from library catalogues and export to a \BibTeX\ {\tt .bib} file, the
approach gives a mechanism for getting information from library
catalogues to \BibTeX.
\item
Alternatively, one might regard the \BibTeX\ database as the ``staging
post''. If a proprietary system exports a \BibTeX\ {\tt .bib} file,
information held in the proprietary database can be converted to a
\BibTeX\ database just before being used in conjunction with \LaTeX.
\end{itemize}
The following problems are likely to arise with such approaches:
\begin{itemize}
\item
The standard ``\LaTeX, \BibTeX, \LaTeX, \LaTeX'' sequence is already
fairly laborious. An additional (``proprietary database to \BibTeX\
database'') stage will make things worse.
\item
Questions could arise about ``which database is the definitive, up-to-date
one\Dash the proprietary system's or \BibTeX's?''.
\item
Mapping problems could arise. The usual \BibTeX\ analysis of
structure (in terms of entry-types and fields) differs from those used
by other systems. In literature-areas where the \BibTeX\ analysis is
relatively coarse, subtleties will be lost if a finer analysis is
mapped to the \BibTeX\ analysis (e.g., if Library Master's public
document, manuscript collection, computer program, audio recording,
video recording, interview, and artwork record-types are all converted
to \verb+@MISC+).
\item
Documentation may be cumbersome, since the end-user will have to
consult that for the proprietary system, that for the conversion
procedure, and that for \BibTeX. The user will also need to
understand the two lots of terminology, and be able to ``translate''
from one to the other.
\end{itemize}
Because of these potential problems, I'm not inclined to pursue this
type of hybrid approach either.
\subsubsection{The user's choice}
Given some {\it modus vivendi}, end-users would be able to make their own
assessments of which bibliographic software suits their needs.
\begin{itemize}
\item
Someone who wants ready-made methods of downloading information from
commercial bibliographic databases, {\SMC CD-ROM}s, library catalogues,
etc., will probably favour one of the proprietary programs. The
proprietary systems also offer database administration and searching
facilities.
\item
Different systems implement different analyses of the structure of
``the literature'' (i.e., using \BibTeX's terminology, there are
different ways of defining entry-types and fields), and
different people also have different viewpoints. E.g., an analysis
that suits a scientist may be too coarse for keeping track of
``primary sources'' in the humanities.
\item Cost is obviously a factor.
\item
An end-user who wants software that has been designed specifically for
use in conjunction with \LaTeX, will probably be inclined to choose
\BibTeX\ or Tib.
\item
\BibTeX's approach makes good use of disk-space. A {\tt .bbl} file
will be smaller than ``near-duplicates of {\tt .tex} files''.
\item
Anyone who does not have the time and patience to deduce (from a
proprietary system's wordprocessor-oriented documentation/menus) what
the \LaTeX-er should do might prefer to wait until someone else has
deduced what is required, and has documented the tricks involved.
\item
The end-user's choice may be constrained by the platform on which they
are using \LaTeX\ (e.g., they may need bibliographic software for a
\UNIX\ system).
\item
Wordprocessor-oriented systems may not support typesetting subtleties
to the degree that \LaTeX-ers would like.
\item
Support (or lack of it) for non-English languages may be another
factor.%
\footnote{Decisions may be needed about whether to try using a proprietary
system's support for diacritics, in the hope of being able to share a
database with colleagues who use wordprocessors. The alternative would be
to have database entries that use \TeX\ encoding for diacritics. Analogous
questions will arise if mathematical or chemical equations appear in
titles.}
\item
End-users may be constrained to use the same system as other people in
their research group (e.g., so that the group can share databases).
\item
Some software supports ``imprecise citations'' (e.g., ``the item in my
database whose author is \dots\ which has \dots\ in the title'').
Others, such as \BibTeX, require a precise citation via a unique key.
People who are continually adding items to their bibliographic
databases may prefer the latter, so as to avoid situations in which a
match becomes ambiguous even though a document's text has not changed.
\end{itemize}
It is unlikely that anyone will find bibliographic software that is perfect
for their needs. However, people are more likely to find something that
suits them if they have a choice than if they have no choice.
\subsection{OR for \LaTeX3}
Given the situation outlined in section \ref{background}, I suggest the
following as the OR for \LaTeX3's relationship with
bibliography-formatting software:
\begin{itemize}
\item
As far as practicable, \LaTeX3 should be neutral towards the
end-user's choice of bibliography-formatting software. Ideally, people
should be able to choose typesetting software for typesetting reasons,
and bibliographic software for bibliographic reasons\Dash their choice
of typesetting software should not restrict their choice of
bibliographic software.
\item
Hence, a {\it modus vivendi\/} between \LaTeX3 and each of the
``main~5'' should be thought up, tested and documented.%
\footnote{The {\it modus vivendi\/} might be along the lines shown in Table
\ref{choice-of-delimiter}, or might be something else that emerges from
practical experience. It doesn't matter much whether the documentation is
provided by the \LaTeX3 project or by the bibliography software vendor,
as long as someone provides it!}
\item
There might be ``a preferred interface'' between \LaTeX3 and
proprietary bibliographic software. If vendors can be persuaded to
support this interface, \LaTeX-ers will get a consistent interface to
proprietary bibliographic software. If not, things will stay
inconsistent (e.g., as shown in Table \ref{choice-of-delimiter}).
\item
In line with the neutrality suggested above, \BibTeX\ will continue to
be supported, but \LaTeX3 documentation will not be particularly
pro-\BibTeX. It is desirable that {\tt .bst} files should be updated
so that \BibTeX\ produces \LaTeX3 commands (designed to satisfy the
requirements listed in section \ref{DIY}) rather than \LaTeX\ 2.09
commands.
\end{itemize}
\section{Miscellaneous}
\label{misc}
\subsection{``Local names'' for keys}
If you are ``doing it yourself'', choice of keys (i.e., in \LaTeX\ 2.09
terms, the arguments for \verb+\bibitem+) is unlikely to be a problem. For
example, you could equally well use {\tt lamport-86} or {\tt latexbook} as
a key for the \LaTeX\ manual. There is no particular need for consistency
from one document to another: you can use {\tt lamport-86} as the key in
one document, and use {\tt latexbook} as the key in another.
However, if you have a large bibliographic database (perhaps shared with a
group of colleagues), it may be impracticable to keep track of keys
assigned on an {\it ad hoc\/} basis, and difficult to guarantee that keys
will stay unique whenever a new item is added to the database.
Moreover, a {\tt .tex} file to be \verb+\input+ may contain bibliographic
details and \LaTeX\ commands that are generated automatically by
bibliographic software (even though \LaTeX\ will have no way of
distinguishing the file from one that you might produce when ``doing it
yourself''). Such bibliographic software might be programmed to assign
keys automatically, e.g.,
\begin{itemize}
\item based on the {\SMC ISBN}, in the case of books
\item of the form {\tt journal-volume-number-page}, in the case of
journal-articles
\item based on ``record number'', if the bibliographic software assigns
a unique number to each record in the database
\item of the form {\tt lamport-86}, constructed automatically from the
``author'' and ``year'' fields in the database.
\end{itemize}
There may be a dilemma about whether to have automatically assigned keys
that are relatively easy to remember, or to have keys that are guaranteed
to stay unique no matter what else gets added to the database.
As an example, consider what key might be used for the \LaTeX\ 2.09 manual:
{\tt lamport-86} is easy-to-remember, but is potentially ambiguous (because
Lamport published other work in 1986);
if the {\SMC ISBN} {\tt 0-201-15790-X} was used as a key, it should stay
unique but would be difficult to remember.
To help cater for such situations, it might be useful if \LaTeX3 allowed
``local names'' for keys, i.e., some mechanism whereby an author could
declare (e.g., in a document's root file) that, for the duration of a
document, a particular ``informal key'' (to be used in in-text citation
commands) should be treated as a synonym for a ``formal key'' (which
appears in an entry in an automatically generated reference-list). For
example, it might be useful to be able to declare that {\tt lamport-86} can
be used as a ``local name'' for {\tt 0-201-15790-X}.
\subsection{Reference-lists that are also indexes}
Another requirement that needs to be borne in mind is for reference-lists
which, as well as providing bibliographic details of sources, provide an
index to the pages on which the sources are cited:
\begin{itemize}
\item
In law books, the requirement is usually for ``front matter'' units
such as ``table of cases'', ``table of statutes'' and ``table of
treaties''. In a typical ``table of cases'', each entry
tells the reader
\begin{itemize}
\item where further details of the case can be found
(e.g., the relevant law report)
\item which pages in the book's main text mention the case.
\end{itemize}
The other types of tables are analogous.
\item
In mainstream academic publications, the requirement will probably be
for a ``combined list of references and author index'' \cite[pp.\ 198
\& 258]{butcher-92}.
\item
Review articles tend to have very large bibliographies. The utility
of such articles might be enhanced if the reader could track a particular
bibliography entry back to the text where it is mentioned.
\end{itemize}
\section{Acknowledgement}
I am grateful to the referees for their comments on a draft of this paper.
The paper incorporates a number of ideas taken from the referees' comments.
\appendix
\section{Some suppliers of mainstream bibliographic software}
\label{suppliers}
\subsection*{EndNote}
Niles and Associates. 2000 Hearst St, Berkeley, CA 94709, USA.
E-mail: {\tt nilesinc@well.sf.ca.us}.
% UK agents: Cherwell Scientific, 27 Park End Street, Oxford OX1 1HU.
% Tel: 0865 794884. A couple of Cherwell people have JANET addresses
% --- {\tt hodgkina@oxford.vax} is one, I think.
\subsection*{Library Master}
Balboa Software, 5845 Yonge St., \POBox 69539,
Willowdale, Ontario, M2M 4K3, Canada. E-mail: {\tt
hahne@epas.utoronto.ca}.
% No UK agent that I know of.
% Current version just does reference-lists. Next version
% should do in-text citations, so they say.
\subsection*{Papyrus}
Research Software Design, 2718 S. W. Kelly St, Suite 181, Portland,
Oregon 97201, USA. E-mail: {\tt RSD@applelink.apple.com}.
% UK agents: Paul Tharagonnet, European Scientific Software Company,
% Britannia Centre, Point Pleasant, Tyne and Wear, NE28 6HQ. Tel:
% 091-295 3000
\subsection*{ProCite}
% The "Inc" cures an underfull hbox!
Personal Bibliographic Software, Inc., \POBox 4250, Ann Arbor,
Michigan 48106, USA. E-mail: \verb+sales@pbsinc.com+ or
\verb+support@pbsinc.com+.
% European Office: PBS. Woodside, Hinksey Hill, Oxford OX1 5AU. Tel:
% 0865 326612. There's a UK user group\Dash I think some official
% (secretary?) is at Birmingham University somewhere (probably in the
% library). I've forgotten the details.
\subsection*{Reference Manager}
Research Information Systems, Camino Corporate Center, 2355
Camino Vida Roble, Carlsbad, CA 92009, USA. E-mail: {\tt
sales@ris.risinc.com}.
% John Cox (Royal Free) thinks the ``best people in the UK'' for RefMan
% stuff are: Bilaney Consultants, St. Julians, Sevenoaks, Kent, TN15
% 0RX. Tel. 0732 450002. Also sold by: Microinfo, P. O. Box 3,
% Omega Park, Alton, Hants, GU34 2PG. Tel: 0420 86848.
\section{E-mail discussion lists about bibliographic software}
\label{discussion-lists}
The {\tt bibsoft} list provides a forum for general discussion
of personal bibliographic database management systems. You can subscribe
by sending a one-line e-mail message of the form\\
{\tt subscribe bibsoft lastname,firstname}\\
to\\
{\tt listserv@indycms.iupui.edu}.
There are also specific discussion lists for EndNote, Library Master and
ProCite. See \cite{stigleman-93}.
The only Papyrus discussion list that I know about is aimed at people in
United Kingdom institutions that have subscribed to a particular (`{\sc
chest}') licensing deal; nevertheless the list seems to act as a general
technical forum. You can subscribe by sending a one-line message of the
form\\
{\tt subscribe chest-papyrus firstname lastname}\\
to\\
{\tt mailbase@mailbase.ac.uk}.
\bibliographystyle{unsrt}
\bibliography{bib-or}
\makesignature
\end{document}