Implement canonical h-card parsing

[aaronpk]

indiewebify.me seems to be assuming the first h-card on the page is the canonical h-card, which is often not the case.

On aaronparecki.com, the first h-card is usually the location of a recent photo.

[tantek]

indiewebify.me needs to implement representative h-card discovery.

See http://microformats.org/wiki/representative-hcard-parsing and think hCard -> h-card while reading etc.

[barnabywalters]

Thanks so much for reporting this bug @aaronpk and for the help fixing it @tantek! The latest version of indiewebify.me handles the following cases:

• A single, representative h-card being found
• Multiple candidates for representative h-cards found, in which case the first is shown with a small warning mentioning that there were multiple
• Single/multiple non-representative h-cards found, in which case the first is shown, with instructions for how to make it representative
• No h-cards found

For example: http://indiewebify.me/validate-h-card/?url=http%3A%2F%2Faaronparecki.com%2F

[aaronpk]

Nicely done! Works great on one of my notes: http://indiewebify.me/validate-h-card/?url=http%3A%2F%2Faaronparecki.com%2Fnotes%2F2014%2F10%2F20%2F4%2F

Does this mean my site should be including a rel=me property of "http://aaronparecki.com/" on every page so that when there are multiple h-cards it knows which to use as the canonical?

[barnabywalters]

@aaronpk seeing as rel=me creates a page-level relation, it’s not really appropriate for use on pages other than the homepage — typically you should be using authorship markup of some sort on pages other than the homepage.

The representative h-card algorithm finds, on pages representing an h-card (but which might have other content on), the h-card that page represents. On pages representing other things, e.g. an event or post, the authorship algorithm is more appropriate.

Currently indiewebify.me only partially implements the authorship algorithm, but with the recent improvements to barnabywalters/mf-cleaner it’ll be trivial to improve this behaviour.