Re: Meaning of header fields Date: and Age:

On Tue, 6 Jul 2004, Koen Holtman wrote:

> On Tue, 6 Jul 2004, Alex Rousskov wrote:
>
> > The age calculation algorithm in RFC 2616 is indeed buggy, but in
> > a different, more subtle, way:
> > http://lists.w3.org/Archives/Public/ietf-http-wg/2002JulSep/0048.html
> >
> > IIRC, nobody commented on that age calculation bug and its
> > proposed fix. It is not mentioned in the RFC 2616 errata. I
> > suspect it will remain with us forever.
>
> I recall that this bug (and ways to fix it) did get discussed back
> in the late 1990s when the spec was being written.  Opinions ranged
> from `it is a bug' to `this is a safe way for the calculation to err
> on the side of caution'.  I recall that the original author of the
> calculation (Jeff Mogul) did intend it this way.  Consensus back
> then was that the bug/feature should be kept in.

Thanks for the info! I may have found that e-mail thread:
http://www.hpl.hp.com/personal/ange/archives/http-wg-archive/0378.html

Jeff Mogul said, "The fact that the response_delay is counted twice in
the final result is an attempt to correct for clock skew, since if you
can't guarantee that the client and server clocks are synchronized
(and we can't!), then the apparent_age could be wrong by a significant
amount."

The formulas I proposed do not use apparent_age or response_delay, so
I cannot apply the above logic directly. Unfortunately, I do not know
what Jeff was using to estimate "clock skew" effect on a formula, so I
am not sure whether the simpler formulas I came up with suffer from
clock skew more than the current complex ones.

As of now, I am only sure that the complexity (and unexpected
results?) of current formulas lead to implementation bugs. I am not
sure whether simpler, natural formulas do not suffer from higher clock
skew sensitivity.

Thanks,

Alex.

Received on Tuesday, 6 July 2004 18:03:04 UTC