ietf
[Top] [All Lists]

Re: Deployment vs the IPv6 community's ambivalence towards large providers

2000-08-22 06:50:02
From: Robert Elz <kre(_at_)munnari(_dot_)OZ(_dot_)AU>

  | Funny, that's EXACTLY what an ATM fan said to me in 1995!

ATM was never rational competition for IP, it competes with FDDI,
ethernet, frame relay, ...

I understand that some ATM advocates thought it could replace everything
but that was never rational.   On the other hand, IPv6 and IPv4 are
essentially the same.   Except IPv6 has enough addresses for everyone.
As soon as there's an application which requires global addresses for
lots of devices (and IP phones just might be that application, if they're
not, something else will be) the continued life of IPv4 with NAT to make
its apparent address space bigger is doomed.

That's unfair to the ATM advocates or too generous to the IPv6 advocates.
The hope that ATM to the desktop would replace IP was a telephant hope
of a dream of a delusion, but consider the parallels:

Both IPv6 and ATM-to-the-desktop-and-replacing-IP
   - claim more host addresses than humanity will ever need
   - claim much better QoS mechanisms
   - involve tunneling IPv4 until the transition is complete
   - require major changes to applications, hosts TCP code, and the boxes
       that connect hosts.
   - claim to be desert toppings AND floor waxes
   - suffer standards committee bloat and silliness, such as the typical
      standards committee doubling of the IPv6 address from 64 to
      128 bits, the lunacy of 48 byte cells and the 54 flavors of AAL.
   - no one yet understands how to really make IPv6 or AAL work or even
       play nice with IPv4 during a transition.  (The big unknowns I see
       with both are in application code.  I have some application code
       that is supposed to do both IPv4 and IPv6, but ....)

The differences are
    - IPv6 is tolerates applications that want datagram-like service
    - some relatively minor technical details such as cell overhead
       and stupid ATM switches that didn't drop entire upper layer
       packets when one cell is lost.
    - ATM switches were supposed to be unlike IPv4 routers, or infinitely
       fast and too cheap need purchase orders because the telcos would
       be buying so many of them, but turned out more like telco CO's.
      IPv6 routers are supposed to be just like IPv4 routers.
    - non-technical forces that might actually cause IPv6 to replace IPv4.
       They can be summarized as IPv6 came from the IETF while ATM was
       from the other guys


Vernon Schryver    vjs(_at_)rhyolite(_dot_)com



<Prev in Thread] Current Thread [Next in Thread>