Judson Valeski wrote:
> Just downloaded and tried it; much better! couple questions:
>
> 1. Did you change the semantics of the reload button; ie. *before*
> were we doing a bypass cache load (I hope not, but want to make sure
> :-)), and we're now doing a load that would leverage the cache?
> Assuming the semantics haven't changed, and that we just have a
> cleaner cache impl now, things are much better. A simple test is
> tinderbox... do a before and after build load of that page and watch
> the mozilla graphic completely reload in the *before* build, and watch
> it snap into place w/ the new build :-).
reload = FORCE_VALIDATE (ignoring prefs, send an If-Modified-Since
header to validate the cached content)
shift-reload = FORCE_RELOAD (ignoring prefs, fetch a fresh copy of the
cached content).
normal browsing = honor cache prefs... ONCE_PER_SESSION really means
potentially use stale content,
whereas FOR_EACH_PAGE means ensure content is valid by for example
comparing the expiration time to the current time and sending an
if-modified-since request if the cached content has expired.
>
> 2. What's in there? new memory cache and new disk cache; or just new
> disk?
HTTP is using a very simple two-file-per-URL disk cache at the moment
(one file for the content and another file for the meta data). Work is
underway to develop a more optimized disk cache.
>
> Jud
>
> Nils Ellmenreich wrote:
>
>> Hi,
>>
>> as Darin Fisher requested feedback in n.p.m.announce for the new cache -
>> here you go. I've been using it for three days now on a RH 6.2 box and
>> since yesterday it's pretty good. Especially
>> http://bugzilla.mozilla.org/show_bug.cgi?id=66940
>> seems to be gone now. Performance on page reloads improved
>> substantially.
>
I believe this is because we are finally calculating expiration times
correctly (according to RFC 2616).
>>
>> One thing that's not quite right: pages with lots of images don't load
>> all of them the first time. If you do a "reload", then you usually get
>> all of them (similar to bug 68151). I've recompiled mozilla with the old
>> cache and the effect is much less visible. Build ID 2001031509, UTC
>> time.
>> (BTW, my cache preference is to compare "once per session").
>
Interesting... can you send me a sample URL that demonstrates this problem.
>>
>> Another thing: frequently changing pages like news tickers don't get the
>> latest page reloaded. Even when I set the cache policy to "every time I
>> view the page", a new visit (per typed URL or bookmark) doesn't refresh
>> the page. Context menu "reload" also doesn't work. Only Shift+Reload
>> button gets the updated page. You might want to look into that.
>
Ok... please use about:cache to inspect the cached content for that
ticker URL... when you pull up about:cache, it is possible to click on
the anchor tag for a particular entry and get HTTP headers. I'd really
appreciate it if you could send me those headers along with the other
info in about:cache for that URL.
>>
>> But thanks for "fixing" 66940 ;-)
>>
>> Cheers, Nils
>
>
Darin
