[ih] ARPANET pioneer Jack Haverty says the internet was never finished

[Brian E Carpenter]

Miles,

You might want to look at
https://www.ietf.org/archive/id/draft-nottingham-avoiding-internet-centralization-02.html
which is discussed at
https://www.ietf.org/mailman/listinfo/architecture-discuss

Regards
    Brian Carpenter

On 06-Mar-22 04:05, Miles Fidelman via Internet-history wrote:
> Jack Haverty via Internet-history wrote:
>>
>> IMHO, if there were a ubiquitous IP-level multicast of some type that
>> could be observed to actually work in the vast reaches of the
>> Internet, people (app developers) who could use it would do so.
>>
>> But "ubiquitous" is important - a mechanism that only works in some
>> places isn't as valuable as one that works everywhere  (a corollary of
>> Metcalfe's Law?).   A mechanism that only exists in one or a 
few ISPs
>> isn't useful unless you expect all your customers to be using that
>> ISP(s), and all of the network paths your customers use (to interact
>> with their own customers etc) are also confined to that same ISP(s)
>> who support the mechanism. Those ISPs of course would need their
>> equipment vendors (routers, switches, hosts, whatever) to also play
>> the same game.
> Exactly.  Which leads to two questions that still remains unclear:
> 
> - If everyone enabled it, how capable, and scalable, is the current
> version of IP multicast?  Judging from my experience with DIS, on the
> Defense Simulation Internet - it can support some very large,
> challenging, real-time training exercises (MMORPGs for folks who use
> real ammo).  But those exercises are one-offs.  A far cry from, say,
> supporting a million videochats.  What are the limits?  Are there any
> clear paths to scaling (if anyone were motivated to)?
> 
> - How much of the lack-of-support is driven by technology, how much be
> administrative complexity, how much by commercial factors?
> 
>>
>> That didn't happen so people invented whatever adhoc mechanisms they
>> needed at some "higher level" where they could just write the code
>> themselves - continuing the "rough consensus and running code", and
>> put their own "servers" (e.g., CDN equipment) wherever it was needed,
>> relying only on the basic unreliable IP datagram delivery service to
>> be ubiquitous.
>>
>> Such "silo-ization" seems to be everywhere now and increasing
>> ....email, messaging, video chat, forums, ....
> 
> I'm not sure that's the primary explanation.
> 
> Seems to me that, back in the day, resource & information sharing were
> the prime drivers for the net - making connectivity and interoperability
> core drivers.  (C.f., Metcalfe's Law).
> 
> Since commercialization of the net, it seems like capturing market share
> has become the fundamental driver - leading to intentional creation of
> walled gardens.  Without much effective pushback.
> 
> I'm reminded of the early days of email:  There was a time when access
> to Internet email was a selling point for Compuserve.  Today, folks are
> selling private email (and chat) based on privacy, codes of conduct (or
> lack thereof), etc.  (Discord doesn't grow because it adds value, it
> grows because it's an alternative to Facebook). We're almost back to the
> days when Boston had a dozen phone companies - each promoting itself
> based on its user base - and every business needing to have a dozen
> phones on the desk.  (Kind of ironic, that Microsoft Exchange supports
> email and calendaring standards better than anybody else.)
> 
> (By the way, not a hypothetical for me, right now - as I'm about to
> launch a new venture that has a major social-networking component.
> Struggling with which standards to build around, and how to gateway to
> other environments - so that we can operate across, and independent of,
> the growing myriad of platforms.)
> 
>>
>> Sigh,
>> Jack
>>
> Sigh, indeed,
> Miles
> :-(
>