--- Comment #6 from Michael Osipov <> ---
(In reply to Mark Thomas from comment #5)
> The proposal never went past draft 03 in 2012. I'm wondering why.

Me too.

> The max parameter is already deprecated in draft 03. I don't think Tomcat
> should be implementing a deprecated feature of a draft proposal without a
> very good reason and I can't see one at this point.

Agreed, I will happily drop max parameter.

> I can see the timeout could be useful in avoiding sending a request just as
> the server was closing the connection, triggering a TCP reset and a need to
> resend the request. However, for that to work the client needs a reasonable
> estimate of the latency between the client and the server and that isn't
> always available.

Not only latency, you can dynamically adjust your connection pool to that and
avoid broken connections. Apache HttpClient does that.

> That it is intended to send this header only when the client sends
> "Connection: keep-alive" doesn't really change things. Browsers usually send
> that. Having to parse the request header and likely generate the response
> header adds overhead to every request. It would be useful to have a sense of
> the scale of that overhead.

The Connection header is parsed anyway with isConnectionClose() already.
Getting an integer from the socket wrapper seems easy. I could at most add
nanoTime before and after and have a look at these numbers, but I doubt that it
will really consume time. WDYT?

You are receiving this mail because:
You are the assignee for the bug.
To unsubscribe, e-mail:
For additional commands, e-mail:

Reply via email to