tomcat-users mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Craig R. McClanahan" <>
Subject Re: Java NIO and IIs/IE ACK/NAK Requests
Date Wed, 08 Jan 2003 06:55:24 GMT

On Wed, 8 Jan 2003, Jacob Hookom wrote:

> Date: Wed, 8 Jan 2003 00:38:22 -0600
> From: Jacob Hookom <>
> Reply-To: Tomcat Users List <>
> To:
> Subject: Java NIO and IIs/IE ACK/NAK Requests
> Two things I'm wondering about:
> First, is there any push to include the new NIO libraries or are they
> already included in the new Tomcat releases; if so, there has been some
> debate as to the actual benefit of them on UNIX based systems pertaining
> to performance over standard IO.

Nobody I know of is pushing it at the moment -- indeed, people whose
opinion I respect have been skeptical of its likely value in a servlet
container.  But if you want to push for it, please speak with code, not
words :-).  Writing a customized version of Coyote that used NIO (without
impacting the core of the servlet container yet) would be the first step
in determining whether it's worth the effort or not.

> Secondly, I was reading a forum post on how IE with IIs servers work by
> pre-acking packets from the server to speed up downloads to the client
> browser.  I know it's just a simple return signal from IIs to accomplish
> this pseudo handshake, but is this a feature that will be integrated
> into Tomcat or Apache down the road?

This would have to be implemented in the JVM (for Tomcat standalone) or in
your webserver (for Tomcat behind a webserver).  There's nothing that
Tomcat can do directly unless it were to ignore the
APIs and implement native code TCP/IP drivers.  I don't know about you,
but *I* am totally uninterested in creating a platform dependency in
Tomcat like this, let alone the fact that it would be quite difficult to
achieve comparable performance via JNI.

Regarding the specifics of what IIS does, deliberately violating the
TCP/IP spec standards (to only assist the browser manufactured by
Microsoft) does not seem like something most Apache HTTPD developers would
be very fond of.  Go ahead and pitch it to them -- but warn me first so
I can stand well back :-).

> -Jacob


To unsubscribe, e-mail:   <>
For additional commands, e-mail: <>

View raw message