Are IRIs valid as HTML attribute values?

[medafor]

Is it valid HTML to use IRIs containing non-ASCII characters as attribute values (e.g. for \[code\]href\[/code\] attributes) instead of URIs? Are there any differences among the HTML flavors (HTML and XHTML, 4 and 5)? At least RFC 3986 seems to imply that it isn't.I realize that it would probably be safer (regarding older and IRI-unaware software) to use percent encoding, but I'm looking for a definitive answer with regards to the standard.So far, I've done some tests with the W3C validator, and unescaped unicode characters in URIs don't trigger any warnings or errors with HTML 4/5 and XHTML 4/5 doctypes (but of course the absence of error messages doesn't imply the absence of errors).At least chrome also supports raw UTF-8 IRIs, but percent-escapes them before firing an HTTP request. Also, my web server (lighttpd) seems to support UTF-8 characters in their percent-encoded as well as in unencoded form in an HTTP request.