Perhaps I can capture it when the original webmention is received and write it to the database. ?
I think I've cracked it! When a webmention is received I should be able to get the details from the comment meta and work from there.
Setting up a new DOMDocument and loading it with the contents of the page that sent the webmention I can loop through the link tags until I find the first one that is an RSS or Atom feed. From here I pull the href attribute and write it back to the comment meta.
Should the feed URL be a relative path (as used by at least one person) I have had to construct the full path based on the root address. Hopefully it will be a reliable process.
Once the feed URL is in the comment meta it can be pulled from the database and inserted into the OPML file as it is constructed.
Now I just need to ensure that the OPML file is properly formatted and that the action to hook into
comment_post actually works properly.
When visiting the Directory the file is recreated dynamically based on the latest webmentions received on the blog. It is filtered to only those with a valid feed (obviously) but no JSON Feed for now.
Maybe I'll include that later once more feed readers support it.
I've had to take a couple of liberties; for example, sites using the Known CMS appear to create an RSS feed on all pages - even a single post - by adding
?_t=rss to the URL. Not much use when it only contains that specific post!
If the feed URL discovered includes this then I change it to
I'm sure I'll come across other peculiarities in future.
Although it may not be 100% up to date you can use the direct link to the OPML - because it is recreated when anyone visits the Directory you get the benefit of newer versions without needing to revisit the Directory yourself.