In Baker's classic: cmos circuit design, layout and simulation, he use 50nm as short channel, and 1um as long channel. But for practical purpose, what is the cutoff line between long channel and short channel device?
I think you should not distinguish between long and short channel devices only by the length of the channel, but by the behavior of the device as a short channel or long channel one, in different conditions. For example, as the view point of velocity saturation, one of the properties that we belong to short channel devices, it is only important to have a vds larger than vdsat and it can be occurred simply in some said long channel devices like 0.25um.
Inside Simon Sze's Semiconductor Device Physics. There is a defination when will be short and when will be long. The effective channel length and electrical field are teh domain one. If from tehcnology bench mark, beow 0.25um it will be easy to get into short channel effect.