Re: What is a "session"?

["John G. Waclawsky" <jgw@cisco.com>]

This is an interesting discussion point. When I think of a browser, which is
what many of us may be thinking when we are considering the "wireless web",
the old concept of a session, used by the OSI model, is merely one of many
“threads” of information flow to get the user a “completed” Web page. In this
sense the “super transaction” of assembling dynamic and static content is what
the user would view as his “transaction” or unit of work. Processing to satisfy
Web page assembly is typically distributed around the network. It begins with
clicking on a URL and usually ends with “Document: Done” being displayed by the
browser. During this time, multiple TCP sessions (from SYN to FIN) are
established and terminated (other traffic flows using UDP could occur as well).
The distributed nature of Web page assembly is clear if you look at trace data.
I hadn't looked, but the statement asserts there is a lot of HTTP state
involved in this process as well. Is this really true? ...or is simply enough
to move all the TCP activity and the browser will take care of things, maybe
simply trigger a page reload and rendering will be take place correctly. Also,
DNS activity is usually recursive and occurring in the background to help
locate web page content and is capable of generating  network traffic
too.         Regards  John

Erik Nordmark wrote:

> draft-reynolds-mobile-isp-requirements-00 talks about session mobility
> without carefully defining a session.
> It lists http as well as ftp as being sessions and goes on to require
> session mobility from one device to another (from the phone to the laptop).
>
> What drives this requirement?
> I think it is rather hard to build the necessary mechanisms (and get
> interoperability) so that you can transfer not only the TCP connection
> state (in order to get ftp session mobility) but also the application
> state (the cookies, sessionids, SSL state and everything else that an http
> browser maintains) so that you can move the http session from the
> (small) browser in the phone to the (larger) browser in the laptop
> and vice versa.
>
> So I'm quite concerned by this requirement as stated.
>
>   Erik