Note that article names allow just about every character so we should probably use _something_ for a text delimiter and create a real api. The api would have the advantage of being able to disable article summaries - they probably only want article titles. That'd be faster.
Version: unspecified
Severity: normal
Another thing - what we mean by multiple pages is to use the text of all the pages as the input to the morelikethis algorithm.
| is safe for a delimiter. |
It's not allowed ( https://en.wikipedia.org/wiki/Foo|bar ), and thus is used in the MediaWiki API for this purpose ( https://en.wikipedia.org/w/api.php?action=query&prop=revisions|info&format=jsonfm&titles=Earth|Venus ).
It could also be used in the text search query (unless there's already another convention for this).
Change 140562 had a related patch set uploaded by Manybubbles:
morelike: with multiple titles