The motivation for 2 XSDs seems to be that it might be burdensome for the
DSMLv2 client/server to consume additional schema that it will not use. If
that's true, then if we were going to optimize the number of XSDs wouldn't we
optimize such that DSMLv2 clients and servers that implemented the "standard"
transports would need only one XSD (that defined just what was needed for
envelopes), rather than optimizing for possible future transports?
I don't see any real reason to pursue that path, however. Given the small
amount of schema that is unique to either
DsmlEnvelopeRequest/DsmlEnvelopeResponse or DsmlRequest/DsmlResponse it seems
prudent to simply define a single XSD for DSMLv2.