begin quoting John H. Robinson, IV as of Tue, May 08, 2007 at 08:33:59AM -0700: > Todd Walton wrote: > > On 5/7/07, James G. Sack (jim) <[EMAIL PROTECTED]> wrote: > > >TCP/IP made explicit the concept of endpoint addresses, and discovered a > > >new definition of "network". > > > > I see, I think. I'm a fifth of the way through the talk and so far > > what I've gotten is that phone numbers are paths. A phone number > > isn't an end point, but rather how to construct a path to the end > > point. Very subtle difference, and I'm not sure I see how it matters.
Thus the concept of "packet-switched" and "circuit-switched" networks. IP is packet-switched. TCP puts circuit-style semantics back on top of the packet-switched IP layer, because the circuit-switched semantics are what we mostly *desire*. "I want to talk to X." The difference is that on a circuit-switched network, the circuit is allocated and reserved for the whole conversation. Bursty conversation, such as one filled with awkward silences, keeps the circuit open, which keeps anyone else from using it. So if you have two awkward conversations and one available wire, you only get to have one conversation with a circuit-switched network, and two conversations with a packet-switched network. To the user, however, it's still a conversation as if over a circuit-switched network. They don't care, unless the packet-switched network degrades performance. > > But something tickles the mind there. Maybe it becomes clearer later. > > That makes me think of the Postman Sort: > http://en.wikipedia.org/wiki/Bucket_sort#Postman.27s_Sort Heh. > A zipcode is not a destination, but it does tell you how to get there. Um, it is a destination... of somewhere *close*. There are places where zip+4 is sufficient. -- Now bang-paths... that tells you how to get there. Stewart Stremler -- [email protected] http://www.kernel-panic.org/cgi-bin/mailman/listinfo/kplug-list
