Antwort: [Mod_gzip] About 'Vary' header insertion

[sara a]

>From: Michael.Schroepl@telekurs.de
>Reply-To: mod_gzip@lists.over.net
>To: mod_gzip@lists.over.net
>Subject: Antwort: [Mod_gzip] About 'Vary' header insertion
>Date: Thu, 10 Oct 2002 01:41:04 +0200
>
>
>Hi Sara,
>
>
> > 2. Adding 'Vary: user-agent, Accept-Encoding' header to it,
> > so that if any caches if they capable, they server the gzipped
> > content correctly based on the user-agent/Accept-Encoding.
>
>If you _want_ the content to be cached, you should encourage
>the client to not ask for revalidations.
>"Expires" and "Cache-Control" are your friends here.
>
> >  Now, the problem, what if the cache does not know about Vary header
> > and discards the header?. Will it end up caching the content?.
>
>You could have asked as well "what if the client doesn't
>know "Pragma". If you assume that any one player in the game
>is broken, then you will be in trouble when predicting the
>behaviour of the whole system.
>
> > But when I use <BACK/FORWARD> button in Netscape 4.7, it re-fetches
> > the page instead of display the previously fetched page
>
>Which caching strategy did you configure in Netscape 4.7?
>And you should know that Netscape believes _very_much_ in
>Expires headers ... sometimes even more than in "Pragma:
>no-cache" for the same response ... ;-)

I am not sure about that. It is just the default installation.

It looks like Netscape 4.7 does not build the history if there is a pragma: 
no-cache in the response.  So, whenever I press <BACK/FORWARD>, it always 
re-fetch the page. I guess, then we should not use pragma: no-cache in the 
response.

>
> > and sometime it displays 'Data not found' error.
>
>Which method are you talking about? GET or POST?
It is a POST method.  For the GET, it silently re-fetch the data from the 
server and for POST it either displays this error message or pops up the 
dialog box asking, for Repost data?

>
> > Now, whats the safer method, to make sure the browser receives
> > compressed response without any hiccups if there are some shared
> > caches on the way.
>
>This depends on numerous aspects you didn't even mention.

Mostly, the dynamic contents are non-cacheable but they are compressible.  
If we take the hit that, compressible response are not cacheable(for safer 
side), then, what are the minimum header required in the response to 
guarentee that no shared-caches in the middle will cache this content.

>After all, you cannot do more than guess who you will have
>to deal with and how this one will comply to the standards-

I agree with you.  Thats why, I am more concern about selecting between 
'cache' or 'no-cache' for the compressed content. I understand the 'Vary' 
header is the best but till every caches implement the Vary in the positive 
manner(either no-cache or perform the validation correctly), we should have 
an alternate.

And, Thanks for your quick response to my posting Micheal.  Your mail is 
really informative.
-Thanks, Sara.

>
>Regards, Michael
>
>
>_______________________________________________
>mod_gzip mailing list
>mod_gzip@lists.over.net
>http://lists.over.net/mailman/listinfo/mod_gzip




_________________________________________________________________
MSN Photos is the easiest way to share and print your photos: 
http://photos.msn.com/support/worldwide.aspx