Liaisons are good. But I have noticed that each of the SDOs I have been
involved in (IEEE, IETF, W3C, OASIS) is very interested in liaising with
the SDOs that were established before it was started and not so interested
in liaising with those that came after. There is also a distinct bias
towards talking to the layers in the stack beyond the one a given SDO works
at,
In short, SDOs tend to look back and look down when they should be looking
forward and looking up.
Right now we have something of a situation with HTTP/2 and QUIC that I
don't think many folk in the HTTP world realize is happening. HTTP was of
course designed as transport for HTML above all else but even from the
start we realized that it also made a decent presentation layer for what
became Web Services, at any ate it was better than anything else around.
HTTP/2 broke that connection, the protocol is 100% optimized for HTML and
Web browser transport. HTTP/2 is not a good transport for Web Services. It
could have tried to be both but the WG didn't want to do that. And that was
probably the right thing to do. Because the needs of Web browsing and Web
services are actually rather different. It is clear to me that Web Services
should look to transition from HTTP/1.1 to whatever QUIC becomes. and that
is not a bad outcome.
My problem is that this is an outcome that we appear to have arrived at by
chance rather than with a dialogue with the stakeholders in the Web
Services world, 95% of which are outside IETF. The principal SDO developing
Web Services is OASIS but we don't seem to be talking to them much or even
acknowledge their existence. And beyond that I am aware of roughly a dozen
more SDOs either being formed, operating or collapsing in the IoT space.
I think that the pairwise model of dialogue isn't scaling. We need to have
some forum at which all the SDOs meet and exchange ideas. So rather than
spending the next three years building QUIC in a corner and then presenting
it as a fait accompli to the Web Service developer community and then
waiting ten years for adoption, get them involved at the start. And rather
than having the design process being decided by one large vendor who nobody
else can dare say boo to, have the other parties round the table and have
skin in the game.
Looking down in the stack, there are features that I really want to see go
away because they have moved up the stack. When Ethernet was first
introduced, all the machines on a network sat on the same coax cable. So
Ethernet absolutely had to have a link layer mechanism for identifying
packet destinations.
Today, networks are rather different. My wired LAN in the house has about a
hundred IP enabled devices on it. But the behavior of that network is very
difficult to predict because much of the routing is happening in stupid
switches at the Ethernet layer rather than at the IP layer which is
actually designed for the task.
I have solid engineering reasons for wanting Ethernet layer routing to just
go away as fast as it possibly can and replace my opaque Ethernet switches
with IP routers.
The problem I have with the idea of 'working together' is that this sort of
change which is natural and obvious is the sort of thing that I suspect
most folk are not going to want to propose in IEEE for fear of causing a
rupture in relations.
And don't get me started on our ICANN entanglements. I have no idea why so
many of you are so keen to invest so much IETF time and political capital
in promoting a project that doesn't actually advantage IETF at all.