Overall, i think we could add something to tokenize the existing external_links field, and then a new field external_link_domains with a reversed field allowing for prefix search with the * at the beginning. I wonder about the use cases, and if perhaps the domain search is the only particularly useful part.

We talked about this a little at the wednesday meeting, a few thoughts:

• It's unclear what the final shape of this should look like, do we want to only match against domain names, or do we want to also match the path portion of urls?
• What about sites like archive.org, should extlinksto:somesite.org find archive.org mirrors of links to that site?
• Should extlinksto:wikipedia.org match aa.wikipedia.org, or should the user be required to enter extlinksto:*.wikipedia.org?

There are probably more questions, but the answers to these questions will inform what kind of text analysis we should apply, and what kind of query should be used. Some ideas (not exhaustive):

• Use pattern tokenizer to extract the domain from the url, match full domains
• Tokenize on . and /, use phrase search to do the matching. This doesn't allow anchoring though, so foo.co would match foo.co.uk.
• Write a custom analysis component into our plugin. This would allow some form of shingling anchored at the TLD and the usage of java url parsing to ensure we parse various unexpected url forms.