On 30/03/2012 7:02 a.m., Tianyin Xu wrote:
> Hi, Amos,
>
> Thanks a lot for the response!!
> The thing I'm still not clear is that it still works when I set the
> limit to "1 bytes" which is obviously less than the size of any error
> message. So, it means any small setting of this directive won't cause
> infinite loop/crash results but only limit the response objects. Am I
> right?
It was fixed a while back, but there are still older versions still
being distributed with some operating systems. For exampel, LTE releases
with 5-10 year support cycles. Being able to use 1 byte limit is a good
sign that your Squid is not affected. The point of the documentation
warning is to make you be careful and check when going small.
>
> This brings the fundamental difference to me. If it may fall into an
> infinite loop, the admin should be really cautious and conservative.
> Otherwise, it doesn't matter too much (at most reject sth) and can be
> set aggressively.
It is reasonable to retain the conservatism regardless of this bug and
test your version before changing these limits. The effects of limiting
responses is clearly visible to the end users, even if this bug is fixed
in your Squid.
Amos
Received on Sun Apr 01 2012 - 02:35:29 MDT
This archive was generated by hypermail 2.2.0 : Sun Apr 01 2012 - 12:00:04 MDT