In 2005, RFC 3986 was the best possible consensus that could have happened at the time. Splitting out IRIs and other scheme definitions was a part of that. The downside of that approach is that RFC 3986 doesn't capture the full definition needed to produce a URI.parse method, and the remaining RFCs never got a critical mass of attention.
Revisiting this in the other direction in 2015 may be in order. Every modern language will have URI.parse as a part of its core runtime library. Everything that is needed to produce such a method should be defined in one specification. New schemes that don't break that definition can certainly be defined, but ones that would break that definition should be pursued in conjunction with an update to the base spec.
In 2005, RFC 3986 was the best possible consensus that could have happened at the time. Splitting out IRIs and other scheme definitions was a part of that. The downside of that approach is that RFC 3986 doesn't capture the full definition needed to produce a URI.parse method, and the remaining RFCs never got a critical mass of attention.
Revisiting this in the other direction in 2015 may be in order. Every modern language will have URI.parse as a part of its core runtime library. Everything that is needed to produce such a method should be defined in one specification. New schemes that don't break that definition can certainly be defined, but ones that would break that definition should be pursued in conjunction with an update to the base spec.
Deciding whether new schemes should be, by default, relative or absolute would be the first step. See: https://www.w3.org/Bugs/Public/show_bug.cgi?id=27233