what's with the stale pages?

Fairly often, when I click on the "earlier" link on my LJ friends or friendsfirends page, it shows me a stale page. I've gotten good at noticing that and hitting reload, but it's annoying. This only happens about one time in five -- most of the time, clicking the link does cause it to reload. Why is it behaving so randomly?

I think maybe I could make this go away by selecting "every time" instead of "when the page is out of date" in Mozilla's cache prefs, but I think that will also make Mozilla go to the network when I hit the Back button, which would suck. I want it to send an If-Modified-Since every time I actually click a link, but use the cached copy for the back button. I think that's how it always used to work. Has something changed?

Is it Mozilla being stupid, or LJ?

Tags: ,

16 Responses:

  1. brad says:

    If you figure it out, do let me know.

  2. nester says:

    IE (ie 6.0xxx, win2k) does not do that, Mozilla (linux) does.

  3. moof says:

    I get the same behavior in phoenix; I never got that behavior with Netscape 4.xx. However, LJ's headers are extremely strange; for two successive HEAD /friends requests:

    HTTP/1.1 200 OK
    Date: Mon, 06 Jan 2003 04:14:51 GMT
    Server: Apache
    Last-Modified: Mon, 06 Jan 2003 04:06:00 GMT
    Cache-Control: private, proxy-revalidate
    Content-length: 37964
    Content-Type: text/html; charset=utf-8
    X-Cache: MISS from http://www.livejournal.com
    Connection: close

    and then:

    HTTP/1.1 200 OK
    Date: Mon, 06 Jan 2003 04:15:43 GMT
    Server: Apache
    Last-Modified: Mon, 06 Jan 2003 04:09:00 GMT
    Cache-Control: private, proxy-revalidate
    Content-length: 38070
    Content-Type: text/html; charset=utf-8
    X-Cache: MISS from http://www.livejournal.com
    Connection: close

    I'm not sure if mozilla/phoenix is getting confused by the out-of-date last-modified times, or what; note that the second request's last-modified time is prior to the first request's Date: field. Maybe some DB server has its clock horribly off.

  4. anthologie says:

    It happens to me in Galeon (Linux) and Netscape (Windows).

  5. communista says:

    LJ's "caching" borks for me in both Opera and IE 6.

  6. hotcrab says:

    happens to me in ie 6 all the time

  7. sarilion says:

    I thought it was a feature? if I'm reading an old thread, but stop for a few hours, I can press earlier fewer times..

  8. crasch says:

    Same thing happens for me with Phoenix on Windows. Konqueror on Linux doesn't seem to do it (though I don't read LJ via that route often).

  9. naturalborn says:

    I think occasionally I have to hit reload on the friends page as well.

  10. brad says:

    A few of the many web servers had NTP running, but apparently not working. Their clocks were off by about a minute.

    It's now fixed. This might be the cause of the problems.

    • thesliver says:

      I just clicked backwards off this page and it brought up the old Friends page and not the current page from where it started. This is using IE 6.026.

  11. reddragdiva says:

    Usually it's Mozilla/Phoenix. Sometimes it's clearly been LJ, though, e.g. when the cache has been cleared.

    I fear I have a lot of use for the prefbar.xpi 'clear cache' button.

  12. ralesk says:

       I fear it's something in the browsers based on Mozilla core.  I found that disabling this "when the page is out of date" thing in the Mozilla cache, it solves it.  Before that, it kept reloading ages old friends?skip pages, as well as once in a while doing me the favour to cache invalid or error report pages.  It was nice to see how /users/ralesk/friends popped up a window in Mozilla telling me how much the site isn't available, while /users/ralesk/friends?skip=25 worked like a charm :]  My other pet peeve is storing 404 and other error pages in the URL suggestions memory...
       And no, jwz, changing it to always look for new versions of the page doesn't -- thank goodness -- change the behaviour of the back button.  It won't look for a new version then, ever. (Unless the crew borks up something, but currently, no...)

  13. teehmar says:

    I've noticed that when it does it, if I view page info on the stale page, it shows an expiration time of something like: 08/18/2016 12:59:32 AM

    Which I'd guess is why Mozilla doesn't refetch it.