- From: Gregory J. Woodhouse <gjw@wnetc.com>
- Date: Fri, 25 Oct 1996 15:53:07 -0700 (PDT)
- To: Jeffrey Mogul <mogul@pa.dec.com>
- Cc: http-wg%cuckoo.hpl.hp.com@hplb.hpl.hp.com
Of course, Jeffrey Mogul is correct that at least in part this really a URI issue in that there is no Content-ID scheme which will allow us to implement caching at this level of granularity. Something off the top of my head (which probably means it has a lot of associated problems) is that complex objects could be encoded using external-body components with HTTP as the access method. Of course, if a component has a URI which references the same host as the origin server it (What if the URI is a URN and the host is unknown? Or for that matter, how many times do we want to query name authorities? It seems we might want to restrict ourselves to URLs here, but I digress) should be possible to use the same connection. Better yet, individual components should be validated upstream if possible (perhaps even on the origin server itself!) --- Gregory Woodhouse gjw@wnetc.com home page: http://www.wnetc.com/ resource page: http://www.wnetc.com/resource/
Received on Friday, 25 October 1996 15:57:59 UTC