RE: Must-revalidate [was Re: Warning: header, need origin]

You're right -- it's an issue of the easiest to understand and implement
encoding. 

Henry's proposal is at least easier to implement when you always obey
the max-age sent by the origin-server, and don't allow preferences for
overriding it.

And I think it may be less likely to confuse -- you basically say that
if the user-preference setting says to not validate when max-age is 0,
then each time max-age=0 is seen, the user has to give premission to
skip the revalidation.

>----------
>From: 	Jeffrey Mogul[SMTP:mogul@pa.dec.com]
>Sent: 	Wednesday, April 10, 1996 3:32 PM
>To: 	Paul Leach
>Cc: 	'http-caching@pa.dec.com'
>Subject: 	Re: Must-revalidate [was Re: Warning: header, need origin] 
>
>    Henry Sanders made an interesting suggestion to me -- what if we
>made
>    max-age=0 always mandatory (same as proposed "must-validate", but
>told
>    people to use max-age=1 when it was (barely) acceptable for
>end-user
>    caches to violate it.
>    
>    Not revalidating when max-age is 0 would require a dialog box each
>time
>    for the user to approve; not validating for max-age > 0 could be
>done
>    with a preference setting.
>    
>    This way, a user-agent that just wanted to strictly obey the
>    origin-server's max-age request could just ignore this whole issue.
>
>I think we (except probably for Roy) are basically nit-picking about
>an encoding scheme.  Henry's proposal seems to be that
>
>   max-age=NNNN
>	(1) what I meant by "must-revalidate" if NNNN == 0
>	
>	(2) what we've already defined for max-age=NNNN for NNNN > 0
>
>I'm not sure if this is any easier (or harder) to implement than
>my proposal.  It won't make the spec a lot shorter (since the
>two different semantics will still have to be explained).  It makes
>the list of cache-control directives one entry shorter.  It makes
>it slightly more likely that someone will be confused about the
>meaning of max-age (since it has this somewhat odd shift).
>
>-Jeff
>

Received on Thursday, 11 April 1996 00:23:11 UTC