Thanks amos.
was very helpful
well if you do ask me i think i know the reasons cause i have seen the
traffic logs
at my work place (some ISP) and some rules sets that people published on
the net.
also i wanted to ask about the squid-appliance plan development.
im not really a developer but it seems like a basic installation script
can be done very easily to configure or\and install proxy with
on and off triggers or basic selection.
also i have seen that Turnkey-linux has a nice "patch" for their core
appliance to install cache and filtering using squid3,
changed easily to other squid versions.
http://www.turnkeylinux.org/forum/general/20100920/tklpatch-web-filter-proxy
a nice thing they have is the TLK config menu based on perl i think that
can be configured to match squid installation\config tool.
their core is 110-150 MB installation footprint gets updates and other
stuff so it seems nice as a candidate.
the only thing i have seen is that my debian as a cache server is using
less cpu and less ram.
I was thinking of taking the time and to try to work on a basic
installation and if i will see that i am managing to make it more than
just installation.
Regards Eliezer
On 23/02/2011 07:56, Amos Jeffries wrote:
On Wed, 23 Feb 2011 07:03:02 +0200, Eliezer wrote:
i have seen refresh_pattern with Age percentage more then 100% and
my question was:
does that percentage does an extending to the expiration time?
or squid has maximum of 100% limit?
No. Limit is 1 year. So if % of past age is over 1 year it will be
cropped back to that 1 year max.
i have seen people writing unreasonable and ridiculously patterns for
cache like:
refresh_pattern -i (get_video\?|videoplayback\?|videodownload\?)
5259487 999% 5259487 override-expire ignore-reload reload-into-ims
ignore-no-cache ignore-private
it means "save the file for 3652(days) = 5259487(minutes)/60/24" am i
right?
Yes they are ridiculous.
Not for the reasons you seem to think.
For an object 10 sec old when Squid received it that % would keep it
in cache and trigger a when it reached 100 seconds old (10sec * 9.99
rounded to 1 second).
The max-valud caps this % but 100 seconds is less than N days, so
nothing happens there.
The min-value then kicks in raises that to "minimum 3652days". Which
is ridiculously long period to go *without validation*.
To get a properly fresh content large % and/.or max-value are
reasonable but such high min-value is usually not a good thing.
Definitely not a good thing to do without deep analysis of the
websites the pattern catches.
no harming anyone but it seems kind of weird.
It is harming their clients view of the websites which match that
refresh_pattern regex. Particularly when those ignore-* and override-*
are used as well.
In the case given it is a youtube (with youtube clone sites as
collateral damage) and a lot of very deep analysis has been done to
ensure that the patterns for those videos does no damage to the user
experience. Quite the opposite. Our adoption and publication of those
rules was a last-resort after a year of discussion attempting to get
youtube to present cache friendly controls on their site fell through.
Amos
