[Wikipedia-l] Please help test new features

[Erik Moeller]

David-
> I think I was unclear about what I meant by stub sections. I meant only
> a section that has just one or two sentences. Some types of articles fit
> into a category and there is a standard set of sections for each article
> in that category, like elements or countries.

This is a fair point, but in these cases of templates I would prefer it if  
authors commented out individual section titles until there is enough text  
to fill them (<!--foo-->). Editors will still see them, but the TOC will  
no longer be triggered.

> If there is an article that has a long first section and a short section
> at the end, say an external links section,  a TOC would be very valuable
> to someone looking for that information. You are probably right that the
> long first section should be broken up, but it is conceivable that there
> could be a long section that is best left intact, like a long quotation
> (say a poem) or a narrative.

We usually do not like long quotations because we are not a source text  
repository. Even if we do have them, we should comment individual parts of  
the quotation, which again allows us to make a logical break. Can you  
explain what you mean with narrative? I would really like to see specific  
examples.

> In fact, it seems that as articles become
> more in-depth articles, longer sections of prose that can't be easily
> broken up will become more common. It seems arbitrary to say "TOCs are
> only for articles with 4 or more headings."

Well, it is less arbitrary than it looks. E.g. Mary Evans writes in "Web  
Design: An Empiricist's Guide":

        Regardless of structure most experts agree that users should not
        have to go through more than three jumps to find the information
        they need.  They added that users will become frustrated if
        presented with more levels of pages acting as tables of contents
        rather than providing content.

Usually, menus should contain about eight items; that is the amount of  
information that most people can process relatively quickly. However, we  
cannot set such a high requirement, of course, because we have shorter  
articles which would nevertheless benefit from a TOC. But if we have only  
two or three items, we can have unforunate results:

For long articles, menu items will be too broad to accurately describe the  
content they describe. Users will find it difficult to use this  
navigational tool except for the most obvious sections (e.g. external  
links). Is the information about Albert Einstein I need in the "Life" or  
in the "Work" part? With so few sections, having no TOC at all would make  
more sense.

For short articles, users will be frustrated because they perceive the TOC  
as another navigation level, which may not have been necessary since the  
information they seek is readily available.

As a general principle, navigation aids should be added only when they are  
truly useful and can save user time; otherwise they can end up being only  
an annoyance. Making tables only visible at four items automates this to  
some extent, although user intelligence is still required to set up the  
sections properly.

> Additionally, we don't want to encourage people to delete or add
> sections unnecessarily just to get a particular desired TOC behavior.

Absolutely, if the TOC behavior is wrong, it either needs to be corrected,  
or as you say, overridden. I'm open to both a "NOTOC" or "FORCETOC" option  
should that ever turn out to be an issue. I'm also focusing mostly on  
Wikipedia's and Wiktionary's use right now and not considering what other  
uses of our software might pop up.

Let's wait how this TOC experiment turns out and how user experiences are  
until drawing any further conclusions. My current belief is that the way  
the TOC is set up, it will encourage positive changes to the way articles  
are structured.

Regards,

Erik