Larry Sanger wrote:

I'd like to see one example of an article, or a class of articles, that would be clearly improved by being put in article-with-subsections form, rather than simply broken into a series of shorter articles.

It might not be that the _articles_ are improved, but the _search engine_ and _navigation in general_. Maybe. :-)

The example I have in mind: Iwo Jima. It's a place. Most people who search for Iwo Jima are not looking for general information about the place, though. Mostly people are interested in Iwo Jima because of the important WWII battle fought there.

A search for Iwo Jima could look like either of these:

WITHOUT SUBSECTIONS [[Iwo Jima]] [[Iwo Jima In World War II]] [[Major Battles of the Pacific Theater in WWII]]

WITH SUBSECTIONS [[Iwo Jima]] [[World War II]] -- subsection on "Iwo Jima as a turning point" [[Major Battles of the Pacific Theater in WWII]] -- subsection on "Iwo Jima"

----

The second and third examples are different. In one case, I am imaginging that the article would actually be written differently. Instead of having a short [[World War II]] article and lots of independent articles, we have a longer article with subsections. There are pros and cons to this.

But in the third case, I'm imagining that the article is written the exact same way in both cases -- but we now have a means to call special attention to the fact that the article does contain a subsection on Iwo Jima, rather than merely _mentioning_ Iwo Jima. This seems a net improvement.

----

One of the mistakes of Nupedia was that we tried to design the perfect system a priori, including grandiose visions of XML-marked-up text to permit all kinds of fancy searching. :-(

At Wikipedia, we take things simpler. But if something is harmless (and ONLY if it is harmless), we might add it.

--Jimbo

I just formulated an interesting hypothesis, which might explain why some people are attracted to ideas like subpages and subsections, despite the many clear disadvantages of them. Sorry, this is going to be a bit of a ramble.

There is an inherent discomfort we all experience when we see that information is not properly contextualized. We don't like to start in the middle of things. General expository works, such as technical books, very often start with very general chapters that present level at a level of generality and simplicity that implies that the rest of the book is going to be a lot simpler than it is; it turns out that, for anyone who can understand the rest of the book, the introduction was old hat and pretty much unnecessary.

In the context of Wikipedia, we'd hate (in varying degrees) to come across an article on the Battle of Iwo Jima that goes into detail about troop movements, generals, casualties, and so forth, and yet doesn't put the thing into the context of the Pacific Theater, which has to be put into a larger context to be properly understood, and so forth. But the contextualizing has to stop somewhere. At some point, we have to let the other articles do the work.

Not quite realizing this, some people come up with this dandy idea of hardwiring the connection between certain subtopics and certain parent topics. This way, nobody can mistake that Mr. Burns is a character in The Simpsons, that Frodo Baggins lives in Middle Earth, and so forth--totally essential, of course, to understanding who Mr. Burns and Frodo are supposed to be. But there are other essential relationships between topics that are equally deserving of being hardwired.

This tendency to want to supply context is *totally* well-meaning and well-intentioned. The problem is that it doesn't neatly solve the problem it is mainly intended to solve, because it reflects a poor understanding of the problem.

The problem subpaging and subsectioning proposals are trying to solve is simply that everything requires context in order to be properly understood. That IS a problem; but subpages and subsections just will not solve that problem by themselves; they have a host of problems themselves, besides.

In hypertext encyclopedia, the contextualization problem is *naturally* and *elegantly* solved with two tools: hyperlinks, and equally importantly, text that is written at a proper level of generality, that *provides* the really *essential* background for a specific topic on the spot. One simply has to distinguish the essential from the nonessential.

An article titled "World War II" simply cannot and should not aspire to be any more than the most superficial introduction to a conflict that requires many volumes just for bare exposition and slenderest theorizing. It should act as a contextualizer of the next level down of subtopics, which might repeat some of the information in "World War II," but which give more detail. And so on, and so on.

Another example is generally given by a number of different articles I've adapted from my introductory philosophy lectures. I *could* have pumped for subsections in order to create, say, eight parts to one long article about ethics. Instead, I used the introductory material, from the front of that section, to try to characterize ethics in general; I then briefly introduced several branches of ethics, such as theory of value and theory of conduct, and created articles for each of the branches. Then, I have not yet but I will further prise out of the articles on the branches (which are themselves very long) more articles about particular ethical theories and concepts, like eudaimonism and eudaimonia.

This requires that we repeat what we say in articles quite a bit. In some of my articles, I have actually cut and pasted bits of text from one article to another. Actually, I think this is a *good* thing. The purpose of encyclopedia articles is to convey human knowledge *on particular subjects* in a way that can be absorbed as easily as the subject allows. So, if I write about ethics and mention what the theory of value is, and then repeat the same thing (in more detail) in the theory of value article, I have done a good job in fulfilling the purpose of encyclopedia articles. If someone wants to know about what ethics (as a Western philosophical discipline) is, it is pretty important that he or she understand at least *what* the theory of value is. One shouldn't have to go to a separate article to find out what the theory of value is, in an article on ethics, because part of what it is to understand what ethics is *just is* to understand what the theory of value is. But an understanding of ethics doesn't require an understanding of all the topics necessary to give a basic introduction to the theory of value. To pick a very specialized topic, you might wonder where to go to find out about Mackie's moral irrealism; that's the sort of detail we might find in an article on irrealist meta-ethics, or an article on Mackie's moral irrealism. Not in an article on ethics, for good reason.

The next part is important: the above implies that there is a distinction between the essential and the nonessential with regard to any topic, and that we focus on the essential when we're writing on the topic, saving the nonessential for more specialized or for tangentially related articles. And *this* is difficult, sometimes. It requires a kind of thinking that mere amassers-of-fact are sometimes not familiar with. I think that just goes to show you that encyclopedias need more than amassers-of-fact. They also need people who really understand the facts and can organize them in articles appropriate to their level of generality--rather than just shoving them willy-nilly into one catch-all article patched together like a Frankenstein monster, with no rhyme or reason. That's sort of what I think of when I think of subsections and subpages. It implies a certain degree of sloppy, disorganized thinking.

But how can we, in practice, distinguish the essential from the nonessential? Like this: we realize that topics in Wikipedia are individuated by their names. One name (or cluster of near synonyms), one topic. Then we ask ourselves: "*Given that* we will also have articles introducing related topics X, Y, and Z, that our readers will be able to consult, and so on down to the most detailed of topics--so that we don't *need* to repeat here all the details of X, Y, and Z--what is really left that's necessary to say in order to understand the meaning or significance of the person, concept, event, etc., right now at hand, and provide the context necessary to understand the articles on the 'subtopics'?"

In writing that article about ethics, for example, I know that we're going to have articles on the theory of value, meta-ethics, applied ethics, and so forth, and it certainly isn't necessary that I repeat the details that I know will be in those articles (or that I repeat the details that will be in even more specialized articles linked to from *those* articles). So I'm off the hook as far as giving a complete, erudite exposition of ethics is concerned. What, then, do I need to say about ethics in order to convey a basic understanding of ethics, *and* provide the basic context necessary to understand the articles on the next level of generality down?

An article on theory of value shouldn't have to repeat, at least in much detail, what ethics is supposed to be (what sort of definitions have been offered, say). So the article on ethics should at least give a definition. Moreover, the theory of value article can probably "rely" on the ethics article to explain some of the basic interrelations between the different branches of ethics.

Larry

On Wed, 13 Feb 2002, Jan Hidders wrote:

From: "Larry Sanger" lsanger@nupedia.com

...

I just formulated an interesting hypothesis, which might explain why some people are attracted to ideas like subpages and subsections, despite the many clear disadvantages of them. Sorry, this is going to be a bit of a ramble. [...]

Hm, I believe you have actually convinced me. Can I summarize you as follows: we don't want people to build their own little hierarchies in this encyclopedia and have some links between them, but we want an interlinked web of relatively self-contained small articles that give a sketch of the context necessary to understand them and link to other articles that describe the context in a general way or parts of the context in more detail.

Correct?

Right! To be clear, it sounds to me like you've summarized what I think people *should* want, and what perhaps many people already do want--but which (I was saying) advocates of subpages and subsections (maybe) don't want. *My* view (to elaborate what you said a bit) is that they should not build their own little hierarchies in this encyclopedia and have *hard-wired* (special, hard-to-remove) links between the elements of them. Instead, as you say, we should have an interlinked web of, indeed, relatively self-contained small (or relatively small) articles that give a sketch of the context necessary to understand them, and link to other articles that describe that context in a more general way or, instead, that describe details of the context in more detail.

Of course, "more general" and "more detailed" are not always neatly distinct or exhaustive categories: for articles that link to each other, there are other important conceptual relationships between their topics.

Larry

On Wed, 13 Feb 2002, Lars Aronsson wrote:

Larry wrote:

...

I'd like to see one example of an article, or a class of articles, that would be clearly improved by being put in article-with-subsections form, rather than simply broken into a series of shorter articles.

I understand that you want an example, and I don't intend to make one.

I understand that you think subsections are a nifty idea, and I don't intend to support the idea. :-) Without an example to work with, discussion seems pretty pointless.

That was why I just mentioned potential power of section headings as loose idea on the tech list. Many encyclopedias have really long articles that are subdivided in this way. Making separate articles (without using subpages) requires that they can stand alone with just a brief introduction at the top. This is fine with [[Swedish monarchs]] or [[History of Poland]], which don't need to be part of [[Sweden]] or [[Poland]]. But "The Middle Ages" is a section heading in [[History of Poland]]. If broken out into a separate article, what title should that page have? Maybe [[Medieval history of Poland]]?

And yet you give an example anyway. :-)

Yes, something like [[medieval history of Poland]] or [[Poland in the Middle Ages]] or even [[medieval Poland]] (by golly!).

There are many features that are "powerful," and yet it would be an extremely bad idea for us to have them. Why? Because they complicate the EDITING system. I'm generally opposed to further complications to the system that would require users to learn much more about how to use the system. If there's one thing we've learned from Nupedia (and it was sure an expensive mistake) it's that not many people aren't going to use a system if it's incredibly complicated. Meanwhile, Wikipedia was very easy to use from the beginning and very quickly became more active than Nupedia.

Again, simplicity is a constraint we have to work with. At some point, it has to care weight with us in designing the system any further. For me, the system is probably *more* complicated than it needs to be.

There's only one really significant change that I would stand behind, that I can think of offhand, which would constitute an addition to the system: *some* sort of categorization system, as we were discussing (probably a multiple-select list of categories next to the "save" button),

This isn't to say that we can't have some very powerful, complicated stuff going on at the metainformation and text processing levels, of course.

It also isn't to say that we can't have some "master version" living somewhere that is beautifully marked up.

Larry