Re: HTTP Expire Header for /check/referer

Nick Kew <nick@webthing.com> wrote:

>CC: ot@w3.mag.keio.ac.jp

Uhm. Why are we CCing Olivier on this all of a sudden?


>On Wed, 10 Jul 2002, Mikael Lind wrote:
>
>>For convenience, I link to the validator using [/check/referer]. The
>>problem with this is that the validator results are cached within my
>>browser [...], meaning that I get outdated results, and somtimes
>>results for the wrong page. Have you considered adding a HTTP Expire
>>header?
>
>Cacheability of validator results is a Good Thing. Unfortunately, in the
>case you describe, it is also a Bad Thing.

Exactly. The results should be cacheable just as any web resource, but for
web authors checking their results, the cacheability combined with poor
behaviour from browsers makes it a problem.

My current thinking on this is to duplicate the Expires information from
the page being validated. This seems a reasonable tradeoff between the two
factors as the author of the original page can most accurately define the
expected change rate of that page. This of course leaves you with the
problem of overriding that behaviour when you're in the middle of a period
of frequent changes to that page.

But the long and short of it is, that I don't think there are any good
solutions to this problem.


>As a point that may be of more academic than practical relevance, the
>correct behaviour of /check/referer would be not to mess about with
>Expires headers, nor even to restrict cacheing with cache-control, but
>to make it clear that content negotiation is happening based on the HTTP
>Referer header.  In other words, it should include "Referer" in a Vary:
>response header.

You obviously are more familiar with HTTP then I am -- gee, isn't that a
surprise :-) -- as that usage never occured to me. I'll read up on it as
soon as I have the chance, and if it makes sense I'll probably add that to
Validator.


>> --
>> I wished for 4 uncursed scrolls of gold detection
>> and all I got was this lousy .signature
>
>On a pedantic note, that's not a valid .sig.

And on a similarly pedantic note, you may want to mention /why/ it's not a
valid signature while you're at it! :-)

Mikael: The standard signature separator is "-- " (dash-dash-space) on a
line by itself. You need that trailing space for it to be a valid signature
separator. If you are using one of the email clients that will munge the
signature separator (*cough*Microsoft Outlook Express*cough*) I suggest you
seriously consider switching to one that won't; as these MUAs tend to have
other problems too.

Received on Thursday, 11 July 2002 02:00:52 UTC