Why is there a confusion out there about ATM and WDM? Why are the two seen as potentially competing, when they should have nothing to do with each other (except perhaps that WDM could be a Layer 1 for ATM, maybe in conjunction with SONET or maybe without SONET). All this confusion exists because the ATM message is very unclear. A lot of people still think of ATM as "a fast backbone technology," which it never was meant to be. That might be SONET, but not ATM. The fact that in MANY circles, potential customers still don't know enough to differentiate between ATM and SONET is inexcusable, in my mind, and an endless source of frustration (for me).
Here would be my strategic plan:
Sell ATM to the RBOCs. That's what the technology was first meant for, and it's still where it makes most sense. Further, if the RBOCs provide ATM to households, it will be hard to stop an ATM avalanche. The RBOCs, of course, need something precisely like ATM in order to make effective use of ADSL and VDSL. There is nothing else so perfectly positioned at this time. And ATM is not nearly as good a fit for any of the other uses it has been targetted towards. Not even close.
As soon as RBOCs provide ATM and ADSL to households, for voice and other services, you will see a large demand for 25 Mb/s ATM products that formerly have had no real excuse to exist. Maybe even slower ATM interfaces would make sense. How about 10 Mb/s ATM?
With households having ATM, now Blockbuster Video stores around the world can each provide movies on demand, over ATM/ADSL, to DTVs (or set-top boxes and regular TVs) inside the home. A digital movie needs 3 Mb/s or better, so a slow ATM over ADSL is perfect for this. And charging for the movie becomes easy. Any form of IP would be iffy at best for this service, but with ATM it's like falling off a log.
The cable comanies, or Blockbuster Video, or new companies, can offer "cable TV" over ATM/ADSL, to DTVs in the home. Dial-up TV, with no fat coax cables anywhere. Use the existing telephone twisted pair already available all over your house. Allow dialing up a channel to watch and perhaps another to record. Charge by the month or by the call, or a zillion other options cable TV over coax does not have.
The ISPs can provide DHCP Internet service as before, but at rates above ISDN. And ATM is perfect for DHCP. The popularity of DHCP mandates that something else exist to identify the IP client, be that a dedicated frequency over a TV cable to a cable modem, or an AESA over ATM, or E.164 over POTS or E.164 over ATM. In other words, DHCP helps ATM in the ISP role. And keeps IPv4 viable that much longer.
What about the backbone? I disagree with Juha Heinanen, actually, that ATM depends on SAR rates above STS-whatever to make it in the future. In the very fast backbones that carry circuit-based ATM comms, you can simply use OC-48 or OC-192, rather than OC-48c and OC-192c. No reason to force yourself to do everything with ATM right away. Use SONET ADMs to upconvert and downconvert the SONET links to rates ATM can handle.
But most of all, there's no reason to think that IP over SONET, sans ATM, should be discouraged. IP over SONET makes good sense for data-only traffic many times, so that's an option ISPs can use. If they want to use ATM to better distribute the packet-switched traffic patterns, that's up to them.
In short, ATM has a role as access method from homes and office desktops and as backbone for the WAN communications that require bandwidth reservations. This is a tremendous role. I'm not sure why everyone can't be happy with this role, and it makes technical sense. Then ATM within office networks, as a way of achieving a seamless connection to the WAN, will also start making sense. No need to bamboozle ourselves with mindless hype.
But all of this hinges on RBOC acceptance. From me, the message would be that ATM will sell itself only if it makes obvious technical sense, and the obvious role for it is from RBOCs to homes and offices. It will grow from there, because that's when it will become the access standard.
Bert (email@example.com) OCT 98