[whatwg] Complex annotations (was Re: Element content models)

[Benjamin Hawkes-Lewis]

I wrote:
> However, I would vehemently stress that it is not that uncommon
> for notes and marginalia to themselves have notes or marginalia,

Then Michael(tm) Smith asked:

> I don't doubt that there are some, but are you aware of any
> specific examples?

Well, most famously (and if you really want your head to spin), have a
look at that masterpiece of hypertext avant la lettre, Pierre
Bayle's /Dictionnaire historique et critique/:

http://www.lib.uchicago.edu/efts/ARTFL/projects/dicos/BAYLE/

And then see how impoverished the experience becomes when translating
into the crude mechanics of current hypertext at:

http://www.pierrebayle.com/

For various reasons (e.g. less prestige in demonstrating familiarity
with the classics; more open reliance on translation; the crassness of
the world of commercial publication; the ready availability of decent
libraries in the rich Western countries), modern scholarship has tended
to have simpler notes.

One major exception however are critical editions. For example, have a
look at Harold Jenkins's magisterial edition of /Hamlet/ (1982). Jenkins
adopts a tripartite scheme of notes. More than half of each page of the
text is filled with notes. Immediately below the text of the play, you
will find notes given collations and emendations of the text. Below that
will you find short discursive notes. Often a discursive note will
include the abbreviation LN, indicating that you should hunt down the
line reference in the back of the book to find Jenkins discoursing on
some subject in exhaustive detail in Longer Notes. (And this is on top
of an introduction longer than the play and an appendix to boot!)

Michael(tm) Smith also raised the problem of how to render such
annotations:

> If you have, for example,numbered footnotes, then how do you number a
> footnote to a footnote? And how would an application programatically
> determine how to number it? If you have marginalia that annotate other
> marginalia, where do you place the additional marginalia?

My considered judgement is that these are all excellent questions to
which we need well-thought out and tested solutions, rather than reasons
to arbitrarily restrict hypertextual annotation to a single dimension.

My less considered suggestion about what those solutions might entail
would be annotation sets which could group (say) "footnotes" and
"endnotes", or "notes" and "long notes". Something like: <note
set="collations"> and <note set="longnotes">.

I don't think browsers would have any great difficult with applying a
default priority of numbering systems. I would note that removing the
numbering of ordered elements from HTML into CSS was a serious error,
because numbering is part of the reference structure of a text,
not /just/ a matter of disposable styles. I think I've even seen legal
documents resort to absurd markup like:

<ul>
<li>2.i The aforementioned agree to [...] </li>
<li>2.ii [...] </li>
</ul>

The form of numbering properly belongs in an attribute, which would be
best inherited from <noteset>, for example:

<noteset id="longnotes" numbering="alphabetic">

If we wanted to provide finer-grain control it would be a good idea to
also allow <note set="longnotes" referenceSymbol="‡">

I also think it would be best if we could have notes that refer outside
of the main document, like XHTML 2 could do something like:

<note set="longnotes" href="longnotes.html#note-d" />

Unfortunately, because of the need for backwards compatibility, HTML5
would presumably have to do something lamer like:

<a rel="note:longnotes" href="longnotes.html#note-d"><note
set="longnotes" /></a>

With regards to positioning, you can put annotations anywhere: inserted
into the text, at the top/bottom/left/right of the viewport, at the end
of the document, in a separate document, in a popup ... there all sorts
of possibilities.

Sorry that these thoughts are a bit rough.

--
Benjamin Hawkes-Lewisw