While others across the industry battle it out over 10-gig claims, Google Fiber is looking to go bigger, plotting the rollout of a symmetrical 20-gig service for select residential and business cus | The service will use Nokia's 25G-PON gear and initially be available later this year as an invite-only product.
In a typical PON (GPON, XG-PON, XGS-PON) you have a single fiber from the central office to the optical splitter in the street, from where up to 64 subscribers are connected one fiber each. The bit between central office and splitter is shared. The splitter is passive and just sends 1/64 of the light to each downstream port, in the other direction it combines all the downsteam light towards the upstream port.
The OLT in the central office sends on one wavelength (e.g. 1577 nm) and all subscriber ONTs send on one other common wavelenth (e.g. 1270 nm). In both directions a time division technique is applied. I believe in the downstream the individual time frames are encrypted with different keys in turn, such that only the specific destination ONT can read the content of their specific time frames. In the upstream the ONTs have to make sure only to send in their own slots, as otherwise the OLT would receive superimposed optical signals that couldn’t be read. You can probably see how this could go wrong if a neighbor had malfunctioning equipment.
The alternative doesn’t really have a set of standards like PON, as you can just use whatever optical transceivers you want for each customer individually. Though I guess that for operational reasons an ISP would still standardise the setup for all customers. For example the ISP whose services I subscribe to tells customers to use “Bidir LR, 10 km, TX1310, RX1490-1550 nm”, as 1G, 10G, or 25G, depending on which you order.
To distringuish such a setup from a PON setup I have seen it being called point-to-point (P2P).