Re: [arch-d] FYI: closure of the IAB Stack Evolution program

[Fernando Gont <[e-mail]>]

On 26/8/19 07:45, Joe Touch wrote:
> 
> 
>> If IP is Internet, and QUIC is transport, what do you call the UDP that
>> stis in between?
> 
> Therein lies the error.
> 
> QUIC alone is not a transport.
> 
> It’s not complete enough to be transport by itself. It doesn’t provide
> the socket (as in RFC793’s idea of a socket) demuxing.
> 
> IMO, QUIC + UDP is a transport layer. A layer that provides a more
> complex service than UDP alone.>
> So yes, UDP is a transport layer - when operating by itself. But when
> paired with QUIC, it is the pair that are the transport.

Maybe one of the underlying issues here is that one would normally
expect one transport service as a single layer (i.e. protocol layer).
But since this is not feasible in the current Internet, you get two
protocol layers to offer a transport service you'd normally implement in
a single one.

i.e., for QUIC you need two protocols to implement a specific transport
service. You use one transport protocol (UDP) that doesn't provide the
service you want (but get the packets through), and then stack abother
protocol on top of it, that does all that is missing.

So I'd argue that it is correct that you can implement new transport
*services*. However, the statement that you can deploy new transport
*protocols* seems to be misleading (you're actually relying on a very
simple transport protocol, and use its services as a building block for
implementing a different transport service). This is what my original
message tried to point out.

-- 
Fernando Gont
SI6 Networks
e-mail: [e-mail]
PGP Fingerprint: 6666 31C6 D484 63B2 8FB1 E3C4 AE25 0D55 1D4E 7492