Keep Alive Questions

dman777 · Veteran Joined: 10 Jan 2007 Posts: 1004

I have a few questions about the keep alive process in detail for keeing a tcp connection on going.

When a client browser sets a keep alive value that is different than a servers keep alive value, which one trumps the other?

What exactly does a keep alive request do and why does the browser send more than 1 during a session?

If the server has a keep alive period of 300 seconds, why would the browser send a keep alive request if it has already been allotted 300 seconds for that tcp connection?

When a web client requests an object such as a css sheet, does that count as a keep alive request?

Veldrin · Posted: Fri Jul 13, 2012 1:13 pm Post subject:

AngelKnight · Tux's lil' helper Joined: 14 Jan 2003 Posts: 127

It's worth being careful about this, because with browsers (as in web browsers, or "user agents" as per the HTTP standard) there's an HTTP keepalive mechanism as well.

How it works is:

The connecting client (or "user agent") sends an HTTP/1.1 request that includes a request header "Connection:". It's usually allowed two values, one of ('keepalive', 'close'). This constitutes a keepalive request on the part of the user agent.
When the HTTP server receives the request, it can elect to refuse a keepalive request, by sending its response including a "Connection: close" header, which signals the client that the server intends to send just one HTTP response on this TCP session, then close it right away.
Or, the HTTP server can decide to send "Connection: keepalive" in the response header, indicating that it chooses to keep the TCP session open with the client in case the client wishes to reuse the session to send another request.
However, the server doesn't indicate how long it intends to keep the session open. It also has the option to close it at any time. Compliant HTTP user agents have to deal with this.