On 8/29/10 11:45 AM, Dahlia Trimble wrote: > This looks to me to be an attempt to provide local caching of relevant > information for reducing lookup requirements and seems a usable > approach as long as the extra data is known to be non-authoritative. > It does bring a implementation-specific form into something that has > potential to becoming a standard of sorts and that may be a concern. > It does not appear to address the problem of fly-by-night service > providers or those who may try to profit excessively from a position > of providing an authority. Personally I don't like the idea of having > a paid subscription service having any control over my ability to surf > the hypergrid, and as such I''d prefer to see the UUID portion be the > authoritative lookup key and the lookup domain portion be considered a > suggestion at best. UUIDs by design have sufficient variability where > collisions are highly unlikely so I don't see collisions as being an > issue even if they are independently generated by independent providers. IMHO, all references should always be to a key type that is by definition not prone to collision. UUIDs were designed for this purpose. That does leave what user name and whatever a UUID maps to. Caching the mapping with something like memcache should be possible. Stuffing possibly mutable [and/or collision prone] data into URIs (effectively pointers) or other identifiers is a bad idea. Identifiers are not supposed to be meaning or semantic bearing.
- s _______________________________________________ Opensim-dev mailing list [email protected] https://lists.berlios.de/mailman/listinfo/opensim-dev
