Opaque vs semi-opaque

Ron Daniel, Jr. writes:
 > It seems to me that 1630 and other documents argue just the opposite -
 > that the semantics of '/', '#', ... are set so that clients can
 > slice and dice the URIs even if they don't know the scheme or
 > namespace. More importantly, the same slicing and dicing code can be
 > used across several known schemes.

I suppose the ambiguity is in what is meant by understanding a scheme.

I really did believe that for any scheme for which the client does
not in fact know that '/' etc can be interpretted, the client should
not do anything with it, i.e. it should treat the URI opaquely.  

But if there are no such schemes because all URI schemes must use
'/' only to represent hierarchy, if they use '/' at all, then clients
always understand at least something about every URI scheme.

If "opaque" doesn't really mean completely opaque, perhaps a different
word should be used to keep people like me from jumping to
conclusions.  (And I didn't even have any caffine today.)

dan