Re: my issues with draft-ietf-tsvwg-udp-lite-01.txt

[Allison Mankin <mankin@psg.com>]

The original proposal was to extend protocol 17 to do both UDP-lite
and UDP.  In fact, the first authors got TSVWG to do a WGLC for two
versions of the draft, one a separate protocol and the other the
shared protocol, to review the pros and cons of the choice.  The
separate protocol won, because it was pointed out that enough
implemenations could suffer from the length field munging by the
checksum coverage stuff if one side knew about UDP-lite and the other
did no.  Another problem was one side expecting the other to deliver
errored data and the other, not a udp-lite speaker, black-holing it
(not caring about the funny length field, because it was one of the
implementations ignoring it and just using IP's length).

As silly as these things sound, the UDP extension was popular in the
open transport community and got a lot of implementation!  People were
taken with the idea of adapting to errored speech delivery with just a
few lines of code in their kernels.

Allison

> > there was, I believe, a proposal for tweaking UDP to do the same thing, so 
> > that the stuff would still be sent with protocol number 21, but with 
> > different semantics. That would be a greater stupidity.
> > 
> > that's no longer on the table.
> > 
> > allison, am I misremembering?
> 
> spent a lot of time staring at packet dumps in octal, have we?
> 
> but yes, i remember this too, whether it really happened or not.