On Mon, Aug 19, 2019 at 01:03:36PM -0400, Keith Moore wrote: > On 8/19/19 11:42 AM, Masataka Ohta wrote: > > But, the most serious, though subtle, flaw killing the Internet is > > in multihoming by routing. > > I wonder if we polled a large number of experts in various aspects of > operating or using the Internet, and asked them "What's the Internet's > biggest flaw?", how many different answers we'd get. More than one answer per participant. > (Note: IETF list is certainly not a representative sample. And IMO there's > not a single "right" answer to this question anyway. But I'm curious about > how broad the spectrum is.) Besides the ietf@xxxxxxxx list itself, my take is that the biggest flaw on the Internet architecture is not so much in it as in the APIs that people built. Specifically the BSD sockets API, which should have had an AF for domainnames from day 0. That weakly implies that DNS should have been there on day 0 as well. It also implies a fair bit about OS architecture. But it's fair to say that the history of BSD and TCP/IP are very closely related. Having bind()/connect() by name would have made address portability less important, and it would have been a much better API anyways than having to call gethostbyname() / getaddrinfo(). Nico --
