I don't like PON either, and I applaud your soapboxing about it, but IMO this overstates the extent of the impending 'rip it all out and replace it'. They can keep most if not all of the fiber runs, and just switch the PON muxes out for DWDM muxes when they need a home run link to each customer.
PON generally uses PLC splitters which are pretty much wavelength agnostic, so you don't even need to swap out the splitters in outside plant. It it entirely possible to overlay DWDM wavelengths on PON segments without even removing or changing any of the PON equipment, making it possible to do a customer by customer migration from PON to DWDM if desired. You do end up having to use 80 or 100km optics to compensate for the insertion loss of the splitter, but it's not like even 10Gbps DWDM optics are too expensive for that (they're on the order of $200 a piece). More important is the security concern as any customers on the PON segment would be able to snoop on traffic making use of MACSEC mandatory.
That said, it is unlikely that major telcos will deploy DWDM to the home outside of niche markets. The savings in feeder fibres costs are nice, but the bigger concern is that there is a very real cost to hosting enough ethernet switches to provide an ethernet port per customer. Most of the GPON deployments around where I live use 1:32 splits, but 1:128 is viable for residential subscribers at shorter distances and when using XGSPON or 10G-EPON (although I stick to 1:32 in my own network). With 48 ports in 1U of space a carrier can serve up to 1536 to 6144 customers in 1U with PON. That would be racks worth of equipment using 1:1 ethernet. DWDM-only would drive up operating costs for space, power, HVAC and equipment maintenance by orders of magnitude.
Yep, you could hack in some DWDM and scale with the capabilities of those endpoints, but at the end of the day it's still running over a shared medium. I don't think it's all impending doom and gloom, just a design decision that I think will not age well. It will be done eventually though I think.
> but at the end of the day it's still running over a shared medium
Everything is eventually a shared medium. You don't have your own fiber all the way to Facebook. So the question is just at which point do you share and that should be a decision made on throughput and cost.
Yeah, as long as your ISP link isn't the bottleneck then it doesn't really matter if they are not as fast as they could be. I'm running on the cheapest FIOS plan and I can count on one hand the number of services where it is the bottleneck. In fact I can only thing of one at the moment: Steam, and even then only sometimes. Even then the difference is downloading a game in 12 minutes instead of 10 minutes assuming it isn't release week on a big game and the servers are slow.
