Wcip enables proxies to cache otherwise non-cacheable objects (by http
cache-control). It also works for cacheable objects, providing better and
no worse cache consistency for the cachable objects, and at much lower cost
to both server and client (via volume validation).
When reverting to http during failure, as long as the proxy doesn't do
adaptive ttl, it's considered adhere to content providers' terms. WCIP
would like to enforce that. However, since proxies are doing adaptive ttl
for any objects without a no-cache directive or expires time, it seems that
the most wcip can ask is that much, i.e., a proxy cannot revert to adaptive
ttl for a no-cache object, but may revert to adaptive ttl for a cacheable
object (although LM may be current time minus 1min and then the ttl would
be small to almost zero anyway).
Content providers have gotten used to proxies doing adaptive ttl to
non-no-cache objects, and they've got tricks to do cache busting. So it
seems there's no value or desire for wcip to say a proxy cannot do it, nor
will proxies conform. WCIP simply lets the content providers set something
that's no-cache or expires in 1min by http terms into cacheable by wcip
terms, while retaining the protection that any non-wcip-conforming cache
will not get to even cache the object or get to cache it for more than 1min.
Thanks!
Dan
This archive was generated by hypermail 2b29 : Thu Nov 18 2004 - 11:22:56 MST