[Standards-JIG] Proposed: Simplified Client Caps

JD Conley jd.conley at coversant.net
Sun Nov 20 22:07:04 UTC 2005


> Chris Mullins a écrit :
> > This eliminates the need for disco requests, serving disco responses,
> > caching, and a number of other complex use cases.
> 
> This turns the registrar into the cache... which doesn't sound sane.
> 

I don't think it turns the regtrar into the cache. Just like we allow extensibility through namespaces we would obviously allow extensibility through ext values. The registrar would just maintain a mapping of popular, standardized, features to ext values. Then in most cases there would be no need for disco at all. For example, if a client advertised "iim" in ext and "iim" happened to be mapped to JEP-0117 (intermediate suite) support then clients could infer a whole slough of features without doing any sort of IQ based discovery. If there are ext's that a client isn't aware of then it would still need to do the disco, but not usually.

Perhaps, to preserve backward compatibility another attribute should be added as "features". This could contain the list of registrar maintained features so that EXT is used solely for "plug-ins" and other "interesting things". Clients which support this caps "features" extension wouldn't need to disco each other at all.

-JD Conley



More information about the Standards mailing list