Re: [squid-users] elapsed time and keepalives (reverse proxy)

From: Henrik Nordstrom <hno@dont-contact.us>
Date: Tue, 23 Apr 2002 18:22:05 +0200

Squid operates at the HTTP level, and as HTTP has no knowledge of the
concept of "pages" Squid does not know.

If you enable log_mime_hdrs then you may be able to deduce which objects
belong together by cross-referencing the Referer: header.. but this is not an
easy task, and also includes objects linked from the page. Browsers do not
tell Squid if the new request is because of the object being inlined in a
"page", or because the user clicked on a link on the page.

The only ones knowing what makes up a page is the one who wrote the page
(person or publishing tool, not the web server), and the browser trying to
display the page.

Regards
Henrik

Eric D. Hendrickson wrote:
> Right, yes I understand that. I am looking for a way to measure
> elapsed time to load a given object (URL), *including* any sub-objects
> (other URLs) within that object.
>
> I recognize that, at the protocol level, they are separate and
> individual requests, but I'd like to be able to group them into a
> "page" and measure the absolute elapsed time for that "page" and any
> "sub-objects" within to load. Not just individually. Of course,
> there is likely to be a lot of overlap since most browsers support
> multiple simultaneous requests...
>
> e.g.
>
> page.html is requested at 1019575895.719 seconds, and completes 13ms
> later. There are two images within that page (separate HTTP requests)
> which each begin at 1019575895.720 (1ms after the first request
> began). One takes 17ms to load and the other takes 18ms. Therefore
> the total elapsed time for this "page" is 19ms.
>
> Is there anything within squid or another way to do this kind of
> measurement?
>
> Thanks, Eric
>
> > A URL identifies a single object, not a whole page.
> >
> > HTTP do not have a notion of "page". HTTP do not at all care how content
> > are linked together, to HTTP it is the same thing if a link is something
> > you click on, or a inlined image or stylesheet, all are simply other
> > URL's linked from the first by it's content. HTTP do not know while
> > processing the first request that it links to other URL's, only when
> > seeing the requests for these other URLs is the linkage revealed to HTTP
> > by the Referer header.
Received on Tue Apr 23 2002 - 10:22:12 MDT

This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 17:07:39 MST