- From: David W. Morris <dwm@shell.portal.com>
- Date: Mon, 8 Jan 1996 15:05:10 -0800 (PST)
- To: HTTP Caching Subgroup <http-caching@pa.dec.com>
On Mon, 8 Jan 1996, Lorenzo Vicisano wrote:
>
> Each server generates its own validators using its own algorithm...
> ...then attaches to validators an ordering prefix.
>
Someone mentioned in past discussion of the Opaque issue that AOL would
route user's to multiple caches which might have different chronological
copies of the same entity. It would be useful, I believe, for there to
be effective order testing of validators.
While it is probably good to bind the ordering algorithm and validator
I would want to avoid any possible chance that:
a. Some process whould attempt to use the ordering algorithm as a
key to making some other use of the validator.
b. That we end up with a proliferation of algorithms such that we
lack widespread implementation and hence interoperability.
I would propose that we might define a single algorithm specified in
an extensible fashion and omissable. The algorithm would require
order based on strict binary comparison of bytes in the stream of bytes
contained in each validator. With a PAD right (or left) rule to
allow for growth. Perhaps two ordering algorithms ... which differ
only in the padding rule for non-equal lenth.
Dave Morris
Received on Monday, 8 January 1996 23:18:51 UTC