this will help ibench performance.
gagan we need this for 0.9.2, this bug is part of the problem for bug 82244, which pav asked me to seperate it out of that main bug
Pav sez that this is not a trivial change and since he has a lot of 0.9.2 bugs already we had moved this out to 0.9.3. If anyone can help us get this in earliar that'll be great!
How much impact does this have on i-bench? See comment from bug 82244 ------- Additional Comments From John Morrison 2001-05-22 22:44 ------- So, I ran the ibench test with the 500MHz win98 PC, and I got: 05/21 05/18 Total 301.1 Total 291.98 First iter. 56.63 First iter. 46.52 Subsq. iter. 34.92 Subsq. iter. 35.07 These are similar to paw's results on the other machine (modulo machine power). I grabbed the server logs from the test server, and after a little slicing and dicing, it turns out that (with post 5/18 builds) we are not caching images properly when they appear in more than one page. e.g., if '/path/to/some.gif' is used in two different pages, it is fetched once for each page, and not from the cache. (I have more detailed data that shows this happening). Now, the zdnet pages serve up a lot of pages that are identical in structure, and which reuse the same graphics page to page. So, effectively, very few of the zdnet pages are in a purely uncached state (some content that they need has been previously pulled in for earlier pages). But, in my test, each page has a unique set of images, so the uncached round is a purely uncached round, not some mixed state.
as what I understand from pav, this is something that if for any reason, a page is reloaded, he is throwing away the images in the image cache, then get a fresh copy from the server, but in fact, should be checking with http validate first and reuse that image in the cache if it's still valid. there is another bug for images not be cached properly between pages, for ibench too... I'll try to find it.
> there is another bug for images not be cached properly between pages, for ibench > too... I'll try to find it. bug 78019
i'm marking this a dup of 108161 since it has patches n stuff in it. *** This bug has been marked as a duplicate of 108161 ***