At 6:04 PM -0700 8/7/08, Tim Bray wrote: >Well, it's not as if the presence of the "http:" scheme requires you >to use the protocol, and in fact a very high proportion of all >accesses to such resources go sideways through various caching schemes >and so on. The notion that the scheme implies the protocol is a >common and very old misconception. > How did you measure the accesses to arrive at the conclusion that a very high proportion of all such accesses to such resources "go sideways through various caching schemes and so on"? Can you elaborate on what "and so on" means here? On your other point, the accuracy of the idea that the scheme "implies the protocol" varies enormously. Some schemes are nearly pure identifier, without strong bindings to a specific dereferencing protocol ( "tag" and "info" come to mind). But some schemes really do have a very strong binding to specific protocol mechanics and a specific URI in those schemes amount to instructions as to how to engage in that. See RFC 4501 for an example. It defines the "dns" uri scheme, using a scheme definition that has a very explicit usage model and that describes the common resolution mechanism using the DNS. It retains the point that other resolution mechanisms are possible, but how to use it in a DNS context is both quite clear and quite clearly understood to be the common case. regards, Ted _______________________________________________ Ietf@xxxxxxxx https://www.ietf.org/mailman/listinfo/ietf