Mailing List Archive

>
> > When I request http://www.apache.org/apache/docs/ the server sends
> > Pragma: nocache
>
> Does hyperreal use a DirectoryIndex of "index"
> instead of "index.html" ?
>
> that would cause the MultiViews code to go look for all
> index.* files, and add a Pragma: no-cache
>
> If you ask for /apache/docs/index.html explicity, you don't
> get the Pragma:

This looks related to some behavior that I have been puzzling
over. ie. includes not getting reincluded when I force a
reload. Instead, the included information is taken from cache.
If I use index.html, it works.

> When I request http://www.apache.org/apache/docs/ the server sends
> Pragma: nocache

Does hyperreal use a DirectoryIndex of "index"
instead of "index.html" ?

that would cause the MultiViews code to go look for all
index.* files, and add a Pragma: no-cache

If you ask for /apache/docs/index.html explicity, you don't
get the Pragma:


robh

>
> >> When I request http://www.apache.org/apache/docs/ the server sends
> >> Pragma: nocache
> >
> >Does hyperreal use a DirectoryIndex of "index"
> >instead of "index.html" ?
> >
> >that would cause the MultiViews code to go look for all
> >index.* files, and add a Pragma: no-cache
>
> Yes, that's it.
> I would suggest that it not set no-cache when there is only one matching file.

that sounds like a good idea.

some worries,

Will it break if index.html (the only match) gets cached, but 5
minutes later, index.cgi is added.

Will a cache flush "index" if it subsequently sees a index.* with
Pragma: no-cache ? (i'd hope so)

-=-=
Are there any spare headers we can send at the moment that caches use to
determine variations, e.g.

index -> index.html or index.cgi with header

Version: html
or
Version: cgi



Roy ?

> > If you ask for /apache/docs/index.html explicity, you don't
> > get the Pragma:
>
> This looks related to some behavior that I have been puzzling
> over. ie. includes not getting reincluded when I force a
> reload. Instead, the included information is taken from cache.
> If I use index.html, it works.

This is a result of E43 and the group x-bit being set on the
file in question.

http://apache.org/httpd/newpatch/bugread.cgi?id=00043

chmod g-x apache/docs/index.html will stop Last-modified: being
added.

It's a feature that you'll get used to. I've used it for months.

rob

On Fri, 21 Apr 1995, David Robinson wrote:
> When I request http://www.apache.org/apache/docs/ the server sends
> Pragma: nocache
> Why?

Because I hadn't turned on "CacheNegotiatedDocs" in httpd.conf (which
isn't in the httpd.conf-dist....) Actually, it was mispelled in the
documentation as "CacheNegotitatedDocs", several times in fact :) Fixed
now.

BTW, my preference would be to *not* have it automatically not send a
Pragma no-cache line when there's only one match, as that could be a
temporary solution. "CacheNegotiatedDocs" is good enough until we can
have the "vary" parameters to returned content.

Brian

--=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=--
brian@organic.com brian@hyperreal.com http://www.[hyperreal,organic].com/

Date: Fri, 21 Apr 95 18:24 BST
From: drtr@ast.cam.ac.uk (David Robinson)

Yes, that's it.
I would suggest that it not set no-cache when there is only one matching file.

David.

Good idea. It shouldn't be too hard to arrange for this; coming this
weekend.

rst

>> When I request http://www.apache.org/apache/docs/ the server sends
>> Pragma: nocache
>
>Does hyperreal use a DirectoryIndex of "index"
>instead of "index.html" ?
>
>that would cause the MultiViews code to go look for all
>index.* files, and add a Pragma: no-cache

Yes, that's it.
I would suggest that it not set no-cache when there is only one matching file.

David.

>> I would suggest that it not set no-cache when there is only one matching
>> file.
>that sounds like a good idea.
>
>some worries,
>
>Will it break if index.html (the only match) gets cached, but 5
>minutes later, index.cgi is added.

That shouldn't matter; caches have to address that problem anyway.
The point of the no-cache was to solve the problem of proxies mapping
diferent requests (with different Accept headers) onto the same cached
object.

>Will a cache flush "index" if it subsequently sees a index.* with
>Pragma: no-cache ? (i'd hope so)

So would I.

David.

> some worries,
>
> Will it break if index.html (the only match) gets cached, but 5
> minutes later, index.cgi is added.

It won't break -- just serve up the wrong document. Caching is not
brain surgery; it doesn't matter if a few bits get handled wrong,
provided that the normal condition is handled correctly.

> Will a cache flush "index" if it subsequently sees a index.* with
> Pragma: no-cache ? (i'd hope so)

No -- how is the cache supposed to know that the two different URLs
represent the same resource?

> -=-=
> Are there any spare headers we can send at the moment that caches use to
> determine variations, e.g.
>
> index -> index.html or index.cgi with header
>
> Version: html
> or
> Version: cgi

Version is already in use. All resources that are intended to vary
based on the client request profile are supposed to include at least
one URI header in the response which indicates the dimensions of the
variance, i.e.:

URI: foo; vary="language,type"

Note that there is no vary value for internal server handlers -- that
wouldn't make any sense to the cache, nor would it care.

......Roy
594 down, 1099 to go