> On 13 May 2019, at 21:28, Brian E Carpenter <brian.e.carpenter@xxxxxxxxx> wrote: > >> On 14-May-19 00:00, Stewart Bryant wrote: >> >> >>> On 13/05/2019 03:49, Melinda Shore wrote: >>>> On 5/12/19 6:39 PM, Brian E Carpenter wrote: >>>> Nor in terms of motivations either. However, it seems relevant to >>>> the question of whether the IETF should continue to do standards >>>> track work *ahead* of the market, of which homenet is a current >>>> example and IPv6 is a 25-year-old example. In the realm of counter- >>>> factuals, imagine for a moment what might have happened if in 1994, >>>> the IETF had said "Too soon, we'll just leave IPng aside for now; >>>> those interested can take it to the IRTF." >> >> Unfortunately there is no such thing as "the market". Different >> stakeholders have different needs in terms of whether they need an RFC >> before they deploy. > > Of course. But what goes into the mass market host operating systems > really matters. Again taking my 25 years old example, it matters that > Windows, MacOS, iOS and Android support dual stack IPv6. That didn't > happen because consumers asked for it. > > Brian > Isn’t the modern world a bit different there. Apps talk to servers across an encrypted tunnel often terminated very close to the user. Inside the tunnel the entire eco system is provided to the app. So the only thing the app needs from the OS is the ability to send raw UDP packets, and the only thing it needs from the network is a UDP path to escape the internet. Within the tunnel the protocol is a private matter between the app and the server and will be dictated by the server provider, or more strictly the server’s cloud provider. Technical domains such as that are often the realm of industry consortia rather than SDOs. This leaves the IETF with the transport net to the server point of presence and the Internet as the network of last resort. Which more or less fits the model of IETF attendance. Stewart
