Re: OS/400 command set

The connection should never be 'idle', Leif.  They only stay connected
when transferring the data from the HTTP server to your web browser.

something like this:

1) Connect
2) Get root web page.  Interpret it while it's downloading...
3) As more images, sounds, applets, frames, etc are found, add them to
   a queue of things to get.
4) Most web browsers will do between 2-5 simultaneous file transfers, so
    if theres anything in the queue, and it doesn't already have N
    transfers running, it'll open a new one...
5) As each session finishes its file transfer session, it gets the next
    file from the queue.  If persistent connections are supported, it can
    do this without disconnecting/reconnecting.
6) When all documents are transferred, all connections are closed.

The only reason you'd have any sort of 'idle' time in any of the
connections would be due to network lag.  So a timeout of 30 seconds is
probably more than reasonable.   The connections aren't left open while
you're viewing the page...  the data is transferred to your PC, put up on
your screen, and the connections are closed.  They are opened again when
you click on something.

You'd CERTAINLY never have one open for days :)


On Thu, 14 Nov 2002, Leif Svalgaard wrote:

> From: Scott Klement <klemscot@klements.com>
> > No, HTTP is not "connectionless".   When you get something from a web
> > server, you connect, you request a document, it sends it, you request
> > the next, it sends it, etc... all during one TCP connection.
>
> The "etc" bothers me. How does it know when my current request is
> going to be the last? The only way this can happen is that the
> connection times out after a (long!) time, because I can sit there
> staring at a webpage for hours before requesting "the next all
> during one TCP connection". Something fishy here....
>