Does it make sense to launch triple play (or double but with data and
tv) service on top of bitstream access? From my research it seems that
such approach lacks control on quality of services (voice, tv, data)
as compared to full local loop unbundling. It seems also that such
solution lacks economic sense. Probably because of this, my analysis
of situation in Europe shows that such services are launchend only by
incumbent operators. But I'm very open to see another point of view or
at least one sucess story or some convincing presentation (bitstream
should be realised on ATM level).

currently in the EU tv over the telephone network is only available on
selected trials so a full impact will not be seen yet however I can
see no reason why its not a good idea. current lines are capable of
taking up to 24mb/s depending on distance from the exchange but its
expected soon to be 8mb/s internet to everyone. assume you have a
2mb/s line and voice taking some bandwidth as is currently the case
this still gives you 6mb/s in which to stream the tv. If you include
the 24mb/s then you have even more space. If they upgrade the system
to rival the cable companies with their 10mb/s internet and tv and
voice that already exist and work well I can see no reason the telecom
industry can not do well at it.

the problem is that in bitstream approach you are dependent
on incumbent infrastructure (in my case atm). that generates
at least three problems:
1. cost - you have to buy bigger stream to serve tv, in normal case
   (no tv) you will get money back charging customer for ip transit.
   i'm not sure that you can charge for tv such fee that will compensate
   buying few megabits just for tv and fraction of it for ip transit.
2. no control - since dslam and atm cloud is not yours you have no 
   control of qos in such network
3. problem partially similiar to previous point. if you for example
   decide to use multicast, you are sending multiple same streams 
   through incumbent's atm network which may be scaled and tailored
   for selling ip data (they for example prepared their network for
   bitstream reselling, but may calculate some high overbooking factors).
   that's why i think full unbundling, in which you will use your own 
   dslam may be better, especially if you make it layer3 device.