The EXT_LINK_URL_CLASS regexp in Parser.php allows the Zs unicode class to delimit autolinked URLs. This includes all unicode "separator, space" character, including non-breaking space (aka \u00A0).
However, it is very common to represent non-breaking space in in wikitext, as it is hard to type the unicode character directly. But doesn't delimit a URL:
http://cscott.net/ this is my website
parses as the url http://cscott.net/ this.
Change 180982 had a related patch set uploaded (by Cscott):
Terminate free external link on (and numeric versions of <>)
My concern here is mostly relating to VE. I was worried that VE would generate in the wikitext if the user manually inserted a non-breaking space, which would then require addition of <nowiki/> to separate it from the URL. If it generates \u00A0 instead, then it looks better (no <nowiki/>) but someone editing in "source mode" can't tell that it's a non-breaking space at all (not so good).
So I think it's better if VE generates and we make the treatment of and \u00A0 consistent in the core parser (and in parsoid).
Change 240568 had a related patch set uploaded (by Cscott):
Terminate autolinks on and numeric entity encodings of <>
Change 180982 merged by jenkins-bot:
Terminate free external link on (and numeric versions of <>)
Change 240568 merged by jenkins-bot:
Terminate autolinks on and numeric entity encodings of <>