From Bugzilla Helper: User-Agent: Mozilla/5.0 (X11; U; Linux 2.4.5 i686; en-US; rv:0.9.1+) Gecko/20010607 BuildID: 2001060708 At http://freshmeat.net and a few other sites, Mozilla 0.9+ displays old content. Hitting the refresh button doesn't refresh the page. Mozilla 0.8.1 doesn't have this problem, and it only happens when using the proxy. Reproducible: Always Steps to Reproduce: 1.Goto http://freshmeat.net and see stale content. 2.Hit refresh button. Actual Results: Old articles are displayed with wrong date and time. Expected Results: Page should have been refreshed with current content. My (Mozilla) memory cache is set to 1024K, and the disk cache is set to 0. My proxy is Squid version 2.3.STABLE4.

Darin, could this be an HTTP validation issue? He has his disk cache set to 0 bytes, which puts it into kind of a strange mode at the moment.

This worksforme with a linux build pulled on 6/11 Reporter: could youtry this with the latest builds

Still having problems with the 6/12/2001 build. Here are the request headers sent to the proxy when I hit reload (Mozilla 0.8.1 shown for comparison): Host: freshmeat.net User-Agent: Mozilla/5.0 (X11; U; Linux 2.4.5 i686; en-US; 0.8.1) Gecko/20010326 Pragma: no-cache Cache-Control: max-age=0 Accept: */* Accept-Language: en Accept-Encoding: gzip,deflate,compress,identity Keep-Alive: 300 Connection: keep-alive Host: freshmeat.net User-Agent: Mozilla/5.0 (X11; U; Linux 2.4.5 i686; en-US; rv:0.9.1+) Gecko/2001$Accept: text/xml, application/xml, application/xhtml+xml, text/html;q=0.9, imag$Accept-Language: en-us Accept-Encoding: gzip,deflate,compress,identity Accept-Charset: ISO-8859-1, utf-8; q=0.667, *; q=0.667 Keep-Alive: 300 Connection: keep-alive If-Modified-Since: Tue, 12 Jun 2001 19:53:13 GMT The newer builds omit the "Pragma: no cache" header and send an If-Modified-Since header instead. Is this correct behavior for a reload?

Reporter: do you see the time displayed on the page changing after you do a reload?

No. The time stays the same after a reload on 0.9+.

benc or tever: could you reproduce this.

it was decided to make reload, reload from the nearest link (in this case the proxy) instead of reload from the origin server. shift-reload forces a reload from the origin server. this behavior (unlike that of moz 0.8.1) is consistent with netscape4x and IE. reporter: please compare mozilla's current behavior to these other browsers. sounds like this bug is INVALID to me.

Mozilla 0.8.1, Netscape 4.76, and IE 5.5 send the "no-cache" directive to the proxy when the reload button is pressed. Mozilla 0.9+ does not. If it was a design decision to reload from the proxy, how about sending a "max- age=0" directive to force the proxy to revalidate its contents? I don't think HTTP/1.0 supports this though.

just testing w/o a proxy server, i don't see IE 5.5 sending the Pragma: no-cache. perhaps it only sends it when talking to a proxy server. nav 4.7 does send the Pragma: no-cache header along with an if-modified-since header. i'm not sure what meaning the if-modified-since header can have if accompanied by a Pragma: no-cache, other than for the origin server. i'm going to try IE with a proxy server.

OK.. so, IE only sends the "Pragma: no-cache" if talking to a proxy server, and it also still sends the conditional headers (If-modified-since and If-none-match). -> me

-qawanted, I've got this +makingtest, I need to research and document this.

this patch returns our behavior to that of mozilla-0.8.1 (and prior versions). according to RFC2616 this is what we should be doing. it was a mistake to have ever changed it. the spec does not require us to send 'Pragma: no-cache' when validating cached content.

r=gagan

let really make sure this gets bang on by qa. (s)r=me

a= asa@mozilla.org for checkin to the trunk. (on behalf of drivers)

fix checked in.

I think the patch broke broke something. On various pages (cnn.com, freshmeat, etc), some images aren't being requested. This problem appeared in today's build (2001061909), and only comes up when using a proxy. Setting the HTTP version to 1.0 gets around the problem. Is this considered a new bug?

If the squid behavior works, I would create a new bug...

sedough: thanks for the follow up info... could you please file a new bug describing the new symptoms/problems. thanks!

It must have been a fluke. The problem disappeared in today's snapshot. Everything works great. Thanks!

Verified.