W3C home > Mailing lists > Public > ietf-http-wg-old@w3.org > September to December 1996

What is a specification for? [was Re: Calculating Age Question]

From: Jeffrey Mogul <mogul@pa.dec.com>
Date: Tue, 26 Nov 96 19:09:40 PST
Message-Id: <9611270309.AA09658@acetes.pa.dec.com>
To: "Roy T. Fielding" <fielding@liege.ICS.UCI.EDU>
Cc: HTTP-WG <http-wg%cuckoo.hpl.hp.com@hplb.hpl.hp.com>
X-Mailing-List: <http-wg@cuckoo.hpl.hp.com> archive/latest/1971
    Regardless of what it says in the spec, the Age value is not touched
    by the cache unless resident_time > 0.

I'm sorry, Roy, but I simply cannot let this go by without comment.

Everyone knows by now that you and I disagree over whether it is
better to overestimate or underestimate the Age.

Everyone also seems to agree on what the specification (i.e.,
draft-ietf-http-v11-spec-07.txt) actually says.  Even you seem
to agree that what the specification SAYS is that the Age calculation
is done in such a way as to tend to overestimate the Age, although
you think this is the technically wrong thing to do.

It may be that, after appropriate discussion in the HTTP working
group, the consensus of the WG is that the next draft of the
specification will change to agree with your desires.  It's
up to the WG chair(s) to declare the consensus on this kind of issue
at the appropriate time.

But, in the absence of a resolved consensus that the latest draft of
the specification is wrong, it is entirely unsupportable to say "I
don't like what is in the specification, so regardless of what the
specification says, I plan to do something different, and I'll tell
other people to violate the language of the specification."

Let's suppose, for example (to choose a MUST at random from 
the HTTP/1.1 spec), that I decided that it is wrong to use
"GMT" as a timezone, and I told people

	Regardless of what it says in the spec, the Date value
	is always represented as an RFC1123 date in Coordinated
	Universal Time (UTC).

You'd presumably give me a hard time about this, and you would
be right to do so.  The standards process has no meaning if people
choose to ignore the plain meaning of the words in a standard.

Having said this, because this specification is a Proposed
Standard, it is perfectly reasonable to experiment with
alternatives, and I assume that both Roy and I would appreciate
seeing the results of a side-by-side comparison of the performance
of several algorithms for computing Age.  But let's not confuse
this with changing the meaning of a specification by fiat.


P.S.: assuming a distinction between "caches" and "proxies"
where the specification makes no such distinction, in order
to find a way to interpret the wording of the specification
according to your desires, doesn't help.  If you think the
spec is wrong, please try to change it, not to create an
external understanding that contradicts the plain language
of the spec.
Received on Tuesday, 26 November 1996 19:36:06 UTC

This archive was generated by hypermail 2.4.0 : Friday, 17 January 2020 17:16:21 UTC