Re: [Spud] updated draft PLUS charter, rev. 1 June

[Joe Touch <[e-mail]>]

On 6/9/2016 9:04 AM, Brian Trammell wrote:
...
> I disagree with the assertion that UDP is actually a transport layer
> protocol. UDP is a facility to allow processes on multitasking
> operating systems to safely send as-raw-as-possible datagrams to each
> other in an environment where one still expects link layer bit errors;
> i.e. the Internet of 1993. Basically all of the functions of a real
> transport layer protocol must be implemented atop UDP-based protocols,
> and most UDP-based application protocols that actually work in the
> Internet either really only send one packet at a time or behave like
> transport-layer protocols.

By your argument, there are no transport protocols.

UDP *is* a transport by providing all-or-nothing messaging, demuxing to
an upper layer protocol/application, and preserving message boundaries.
Granted that's not everything some apps want, but neither is TCP.

Consider that TCP provides reliable, ordered, bytestream service even
though many apps really want message boundaries. Or concurrent
interleaved messages. Or message chunking-and-muxing to avoid HOL
blocking of large messages. I.e., that's how we ended up with a lot of
what's in HTTP - to adjust for things TCP didn't provide or to undo some
things it did provide.

Your argument that apps often want services existing transports don't
exactly provide is undeniable, but that doesn't mean these aren't
transport protocols.

You've only pointed out the known flaw in the OSI model - that there's
no such thing as a fixed number of protocol layers or any one layer with
a single defining property.

Joe