Henrik,
Thanks for your prompt response.
My comments/additional questions are embedded with >>SK.
Thanks,
Steve.
On Fri, 02 Jun 2006 11:10:41 +0200
Henrik Nordstrom <henrik@xxxxxxxxxxxxxxxxxxx> wrote:
tor 2006-06-01 klockan 13:43 -0400 skrev
sknipe@xxxxxxxxxx:
However, our html pages may change in the expires time
frame.
Then you need to advertise a smaller expires interval or
use
"Cache-Control: max-age=XXX" to tell caches they must
validate the
page..
Perhaps a better model for us is more of an update based
model using ETag. The down side of this is that our
infrastructure is hit for each request. However, the hit
is minimal if the html page has not changed.
In this scenario ETag just allows for a stronger
"If-Modified-Since". In
terms of caching of simple responses (no Vary) there is
not much
difference if you use the ETag or Last-Modified. And
with both you need
to set freshness appropriately to have the page
revalidated with the
frequency you need.
Also, I noticed that Squid forces http communication
into
http 1.0 mode. This has a negative effect when working
with IE and the ETag model. Can we have Squid interact
in
http 1.1?
SK The following HTTP response is generated by squid for IE.
Response Header IE:
(Status-Line) HTTP/1.0 200 OK (Please note the HTTP/1.0)
Response Header FireFox:
(Status Line) HTTP/1.x 200 OK (The 1.x works fine.)
In this case, IE will not send back the if-none-match with
the appropriate ETag.
Could you please verify the following. When ETag is used
for .html pages, the caching seems to be at the Browser
level. Does Squid play a role in the if-none-match with
ETags?
Is there a way to have an HTTP/1.1 or HTTP/1.x response
generated for IE?
Thanks very much,
Steve.
Can you please elaborate on what negative effects is
seen with IE here..
For the image, css and js files, we have a requirement
to
check if users are authorized to files before they are
serviced. Therefore, once cached in Squid they are
serviced by Squid and do not interface with our
infrastructure. Is there a mechanism in Squid to enable
the appropriate authorization checks before servicing?
"Cache-Control: s-maxage=0" forces shared caches to
revalidate the
request with the origin server on every request.
Please note that there generally is not much benefit to
cache smallish
objects (<10KB or so) which needs to be revalidated. The
main benefit of
caching small objects is the latency improvements, and
these are almost
completely lost if the object needs to be revalidated..
Regards
Henrik
