Link differently to classes, title properties, capabilities

[kba]

So far, we're using the auto-generated id= attributes of the headings for internal linking, which is great for linking to "non-normative" sections but not ideal when referring to specific terms like classes (see #46).

We're using bikeshed for building the HTML spec, which is used by both WHATWG and W3C for various specs like CSS and HTML, so it comes with a lot of shortcuts for defining and linking to various types of terms, values, interfaces etc. To make use of this, we need to map terminologies:

• Classes define the one hOCR type of an element, so for all intents and purposes, they are equivalent to HTML elements. From hOCR perspective, It doesn't matter whether an element has tagname div or span or p but that it has exactly one class attribute that starts with ocr.
• Title properties describe the layout/typography/recognition-related attributes of an element in key-value pairs. Most of those can be mapped to CSS properties, so treating them like property/value terms seems the best option.
• Metadata are properties of the whole document, again key-value-pairs, so property/value fits
• Capabilities can be classes or predefined values starting with ocrp. The latter are specific values for the metadata property ocr-capabilities.
• Profiles: This is so underspecified at the moment that I skipped them for now.

To reference terms within the specs, I'd propose this scheme:

hOCR terminology	CSS terminology	Definition	Link
class	element	<dfn element>ocr_page</dfn>	<{ocr_page}>
title property	property	<dfn property>bbox</dfn>	'bbox'
metadata	property	<dfn property>ocr-system</dfn>	'ocr-system'
capability	property value	<dfn for="ocr-capabilities">ocrp_lang</dfn>	''ocr-capabilities/ocrp_lang''

See http://kba.github.io/test/hocr-relink/ for a preview. Notice the list of terms and the backreferences if you click e.g. on http://kba.github.io/test/hocr-relink/#bbox.

[amitdo]

Nice. :)

I think a section should not reference itself (See ocr_carea).

[kba]

I think a section should not reference itself (See ocr_carea).

It's not, it's referring to the element which is defined in the section heading. For consistency I would leave it as it is.

I'll go ahead and merge this now to avoid conflicts, since the changes are all over the text. I'll integrate the classication above into the text.