[ih] A revolution in Internet point-of-view - Was Re: Internet analyses (Was Re: IPv8...)

[Brian E Carpenter]

John,

On 30-Apr-26 12:04, John Day via Internet-history wrote:
> Choosing TCP. 

That was before the IETF existed.

> Choosing SNMP over HEMS.

[I can't comment as I wasn't involved at all.]

> Choosing IPng over IPv7, 

If by "IPv7" you mean TUBA/CLNP, there was strong advocacy for it within the IPng effort (from me among others), as well as a dedicated IETF WG, but the rough consensus was for SIPP (which mutated into IPv6). It was very definitely a community process and the consensus was pretty rough. There was no external direction except for political pressure in favour of OSI (in particular the US GOSIP profile), so the SIPP/IPv6 choice was _against_ the only external pressure that I was aware of.

If by "IPv7" you mean TP/IX, a.k.a. CATNIP, the answer is similar, but CATNIP didn't have many advocates, unlike TUBA.


> I am probably wrong but was the choice of domain-names broadly decided. There was a real fascination at the time with ‘host-names’, even though it has been known since the early 80s that ‘host-hames’ are irrelevant to creating end-to-end connections.

[I'm not sure what you're asking there, but if it's DNS, I can't comment as I wasn't involved at all.]

> 
> As long as we are on the topic, why was IP Fragmentation never solved?  PMTUD is a kludge, not a solution. (There is a solution). It is even richer that IPv6 has made PMTUD mandatory, even though it is a known DOS attack method, a real mark of a successful project.

As you know, fragmentation is handled differently in v6 (i.e. only the sending host can fragment a packet) but I think most people agree that we'd be better off without it, or rather, have it handled exclusively by layer 2.

> 
> As for just plain conservative choices: 

Yes, IPv6 was consciously a very conservative design.

> TCP/IP, UDP, losing the Internet Layer, Sockets, TCP Congestion Control was a step back to CUTE+AIMD, IPv6, naming the interface in routers (apparently not understanding that when a packet is sent down a point-to-point line there is only one place for it to come out: the other end. Addresses on either end or not necessary. An identifier local to the router is required to distinguish them but not an address.)

Which is why unnumbered links are a thing.

    Brian

> Most routers are connected by point-to-point lines. But that is okay, routers assigned an IP address to the Loopback module and route on that.
> 
> John
> 
>> On Apr 29, 2026, at 19:47, Dave Crocker <dhc at dcrocker.net> wrote:
>>
>> On 4/29/2026 4:43 PM, John Day wrote:
>>> When presented with potential solutions, the IETF was either directed to take a direction (usually a step backwards), or chose the more conservative decision (also a step backwards) and flawed direction.
>>
>>
>> John,
>>
>> Please provide specific examples of the IETF being given direction as to what choices to make.
>>
>> d/
>>
>> -- 
>> Dave Crocker
>>
>> dhc at dcrocker.net
>> bluesky: @dcrocker.bsky.social
>> mast: @dcrocker at mastodon.social
>> +1.408.329.0791
>>
>> Volunteer, Silicon Valley Chapter
>> Northern California Coastal Region
>> Information & Planning Coordinator
>> American Red Cross
>> dave.crocker2 at redcross.org
>>
>