https://bz.apache.org/bugzilla/show_bug.cgi?id=57921

--- Comment #11 from Rainer Jung <[email protected]> ---
I'm going to change the title of this issue to describe more precisely the
original problem. To discuss further optimizations we can use the dev list or
another specific ticket.

I don't think the problem is that big. The most common case that after applying
the fix for this issue here (HTTP/1.1 with no Connection keep-alive header
leads to infinite keep alive behavior) IMHO remains is the possible race
between the client sending a followon request very shortly before the
keep-alive ends and arriving at the server shortly after. This race could be
made rare by subtracting a small delta from the keep-alive timeout determined
by a connection header.

Another case would be a user who configures a fixed keep-alive timeout using
IDLE_STRATEGY which is longer than what the server supports. That would then be
a user error.

Finally a server could announce some timeout but then due to increasing load
decide to close a connection earlier than announced. That indeed could be a
case for RetryExec.

-- 
You are receiving this mail because:
You are the assignee for the bug.

Reply via email to