Path: utzoo!mnetor!tmsoft!torsqnt!news-server.csri.toronto.edu!cs.utexas.edu!uunet!shelby!neon!opbibtex From: opbibtex@Neon.Stanford.EDU (Oren Patashnik) Newsgroups: comp.text.tex Subject: Re: sorting in bibtex Message-ID: <1991Feb13.151317.21497@Neon.Stanford.EDU> Date: 13 Feb 91 15:13:17 GMT References: <1991Feb7.175558.9848@Neon.Stanford.EDU> <1991Feb9.203132.6226@Neon.Stanford.EDU> Organization: Computer Science Department, Stanford University Lines: 150 [This article is fairly long, because I address not only the specific issues that Sebastian raises, but also some related issues that others have raised in private e-mail discussions with me, or that people have asked me to post.] In article spqr@ecs.soton.ac.uk (Sebastian Rahtz) writes: > In article [. . .] opbibtex@Neon.Stanford.EDU (Oren Patashnik) writes: > > By contrast, author-date styles encourage flabby writing. For > instance many such styles almost require the passive voice---"It has > been shown [Knuth 76] that ..."; the `plain' style avoids the passive > voice---"Knuth [13] shows that ..." > > The usage `it has been shown (Knuth 1987)' is often found in BibTeX > usage because of the difficulty of getting \cite{Knuth/87} to produce > (1987), as in `Knuth has shown (1987)'; yes, I know about \shortcite, > but that's not distributed as standard with BibTeX. You've raised, directly or implicitly, several points in that sentence. What I perceive to be your main point---that people who use BibTeX tend to use the form `(name year)' more often than the form `(year)' because the `(year)' form doesn't come standard with BibTeX ---has a germ of truth, but is inaccurate on two counts. First, you've incorrectly implied that the `(name year)' form *does* come standard with BibTeX. There are only four standard styles (plain, abbrv, unsrt, alpha) and none of them uses such a form. There have been several styles written that use that form (except that most of those styles put a comma between the name and year); I even wrote one of them myself (apalike). Some such styles live in certain repositories, and some may even come on certain distribution tapes, but as of the .99 version of BibTeX, only the four styles mentioned above come standard with BibTeX. Your real complaint, I think, is that it's easier to get your hands on a style that uses `(name, year)' than on one using `(year)' alone. Which leads to the second inaccuracy: I think you've put the cart before the horse. It's not that more BibTeX users use the `(name, year)' form because that's the form that's easier to find; rather, the `(name, year)' form is easier to find because there's been more of a demand for that form, hence that's the form that style writers have produced. (At least that demand explains why apalike---perhaps the most easily accessible such style---was written with the `(name, year)' form only.) But there's nothing inherent in either BibTeX or LaTeX that makes the `(year)' form harder to write a style for; in fact it's a little easier. Incidentally, you've implicitly stated that you prefer the `(year)' form to the `(name, year)' form. I do too; I think the `(year)' form, in the text at least, avoids many of the problems of the `(name, year)' form. But most of its reference-list logical deficiencies (mentioned in my previous article) remain. You brought up another issue: Which styles are standardly distributed with BibTeX? Here's a slightly oversimplified view of the TeX/LaTeX/ BibTeX world. With TeX, Knuth provided a basic product (TeX the program) along with a useful macro package (plain); with LaTeX, Lamport provided a higher-level macro package, along with some useful styles (article, report, book). Similarly, with BibTeX there's a basic product (BibTeX the program) and sample formats---one well- thought-out style (plain) and several useful variations (abbrv, unsrt, alpha). The variations were chosen because they used reasonably common features and hence gave examples of typical functions that one might program in the bibliography-style (.bst) language. So for versions .98 and .99, there were just those four standard styles. Incidentally, they are called `standard styles' because they are standardly distributed with BibTeX, not because they are thought to implement somebody's standards (even though the `plain' style does come pretty close to the style recommended by van Leunen in her "Handbook for Scholars"). For BibTeX version 1.00, which will be the frozen version (in the same way that TeX 3.0 was a frozen version) there will be a few changes. In addition to the four standard styles, there will probably be the four semi-standard styles, acm, apalike, ieeetr, siam, which are four other styles that I've written or maintained; they will (probably) be distributed with BibTeX because they provide more examples (and because it will make my life easier to have all the styles I maintain distributed together), but not because there's something special about them. (By the way, current plans are to add to apalike the `(year)' form, along with a few other APA recommendations, and perhaps change the name to simply `apa' if I decide that what results is close enough to the APA style.) > I find it much easier to follow the argument in `Knuth has so often > shown (1983, 1984, 1990)' than in `Knuth has so often shown (13, 67, > 89)', because the year references provide an intermediate clue about > whether i want to bother following-up the reference. I think the year serves two purposes in the style you prefer: Its primary purpose is to provide a pointer into the reference list (you might need to distinguish among several works by Knuth, hence you need a `1983'---you might even need a `1983a' and a `1983b' to distinguish between two of his works from the same year); but as I argued in my previous article, it's inferior, as a pointer into the reference list, to a number in brackets. As an often useful secondary purpose, it also provides, as you say, an intermediate clue (additional information) about the particular work; but why force the writer to always give the year as an intermediate clue---why not let the writer choose the information, if any, that's appropriate for the sentence at hand? One answer is that many writers are negligent and, by habit, never provide any additional information; it's better to force them, through the citation style, to at least provide the year. That view has some merit; but I think it would be better if writers would get in the habit of thinking about the information they are providing the reader, and if journal editors would encourage such thinking through their editorial policy. [Here's one simply stated rule that should, I think, be included in editorial policies: Always use the number-in- brackets as a parenthetical remark (`. . . as Knuth [13] shows') rather than as a part of speech (`. . . as shown in [13]').] Perhaps there will be a trend in that editorial direction once people realize that the main reason that author-date styles have become popular is no longer valid (as I pointed out in my previous posting, in quoting the Chicago Manual of Style). To sum up: I think it's better to decouple the two purposes served by the `(year)' form, so that each purpose may be served better; the `plain' style does that. So while it's true that a year conveys more information than a number in brackets, it does so by serving two purposes; the number in brackets, in serving just one (reference-list pointer) purpose, lets the writer provide whatever information is necessary to serve the other purpose, in a more flexible way. > I like very much to have instant feedback on the date of a reference, > and I also like the fact that, in a field with which I am familiar, I > recognize `Knuth 1986' as "oh yes, that paper, I've read that", > without having to bother flick through the bibliography. I suspect that what you really want is not instant feedback on the date of reference, but rather instant feedback on which work the author is referring to; the date of reference happens to be one way to try to give that feedback. But I claim that, in a field with which you are familiar, `Knuth [13] shows that . . .', together with the information contained in the `. . .', almost surely serves to uniquely identify the work. One last point. I'm not claiming that there's no utility in seeing the year of publication. To the contrary, I agree that a name and year pinpoint the reference pretty quickly for a reader who's familiar with the field and who gets used to seeing that name-year combination in paper after paper. But by the same token, if you rely solely on that name-year convention, you're putting at an unnecessary disadvantage the reader who is *unfamiliar* with the field: You're quite possibly omitting some information that would be as useful to the reader familiar with the field as the year information is, but that would be much more useful than the year information is to the reader who's unfamiliar with the field. (If you don't rely solely on the name-year convention---for example if you take a well-written sentence in the `plain' style and simply substitute a year for a number in brackets---many of my objections about the text itself disappear, although, again, the reference-list deficiencies remain.) --Oren Patashnik (opbibtex@neon.stanford.edu)