About regexp library choise, I'd say the support for wide charsets
should not only be considered, but required. Regex++ supports it, that's
all I know for now.

about substring vs regex discussion, I'd like to remember regex is not what was proposed initially (regex was only a subset of this proposition). The previous proposition was something like that
( N =~ avi$ || N =~ ogm$ ) && ( T =~ ^video/ ) && ( S > 450000000 )
which can natively mix both regex and substring.

Eric