Re: [nat66] NPTv6 deals with "packets", not with "datagrams" - to be corrected after draft-14

[Fred Baker <[e-mail]>]

On May 3, 2011, at 1:38 AM, Rémi Després wrote:
>> I also am confused by your use of the term "datagram". A datagram is not a transport layer construct,
> 
> Isn't User Datagram Protocol a transport layer protocol?

Yes, and unless the datagram is carried in IP, it's not actually a datagram. I refer you to the definition of the term. A datagram is self-addressed and contains all of the information necessary to deliver it from its source to its destination.

>> it's a network layer construct. http://dictionary.reference.com/browse/datagram
> 
> To know what a datagram is in IETF, RFC's are IMHO better than an online dictionary

If you want to play that game, so be it. The word "datagram" (along with "catenet", which has largely been replaced by the term "internet", in lwer case) was common in the 1970's and 1980's, and the definition was well known to those present. So I don't find a place that says "when I say 'datagram' I mean '...'" in the RFC series - that was in published conference papers circa 1972 when the datagram model was first being proposed as an alternative to virtual circuits. What I do find, however, is this in IEN #48. The Internet Engineering Notes (IENs) were  a parallel document track used in the 1970's. IEN #48 is "The Catenet Model for Internetworking", written by Vint Cerf. It says, among its assumptions, that 

	it is assumed that the
	participating network allows switched access so that any source
	computer can quickly enter datagrams for successive and different
	destination computers with little or no delay (i.e., on the order
	of tens of milliseconds or less switching time).

	Under these assumptions, we can readily include networks which
	offer "datagram" interfaces to subscribing host computers.  That
	is, the switching is done by the network based on a destination
	address contained in each datagram passing across the host to
	network interface.

A datagram is, unlike a packet carried in a virtual circuit, a message that carries its own destination address and can therefore be routed by the network without setting up such a circuit first.

> I hope the IESG won't accept your new interpretation.

Well, if they have a problem with it, they can ask the designers of the architecture for their opinions. "Datagram" is a well known and well understood term.