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