Re: Comments on Byte range draft

Jeffrey Mogul writes:
 >     (optional) (3) we extend the set of headers to include the
 >     generalization "validator" which might be used to contain (e.g.) a
 >     date, filesize, hash, checksum, or random number chosen by the server
 >     as pertaining to a unique version of the document, and allow
 >     "validator=<xxxxx>" to appear as a header in server responses as well
 >     as in this header.
 > 
 > As much as it pains me to disagree with someone who is trying to
 > agree with me ...
 > 
 > The most important feature of a cache-validator is that it is 
 > 100.00000000% opaque to clients and caches.  Checksums aren't
 > opaque; dates aren't opaque; hashes aren't opaque.
 > 
 > If the client or cache is able to do anything with the validator
 > besides send it back to the server, then it won't work.  Which
 > is to say that some creative but misguided programmer will try
 > to use it in a way that will, sooner or later, make us all unhappy.
 > 
 > -Jeff

Much as I hate to agree with someone who is trying to disagree with
me, I hate to disappoint you, but I think we're still agreeing.  My
suggestions were intended as things the server could use in computing
the validator, not things the client would *know* were in the
validator.  I certainly DID NOT mean to imply that strings such as

	"Validator:date=01-March-1995" or "Validator:checksum=19283719827"

would appear.  I meant we'd see

	"Validator:8923984792"  and that the server would be

responsible for generating and understanding this.  Sorry for being,
uh, opaque about this.

--Shel

Received on Monday, 13 November 1995 20:15:24 UTC