IMO, Lens IS makes more sense, as the IS is suited for individual lens, due to the different focal length, apeture size etc etc.
although it makes more economy sense to build IS into the body as you pay for it once only.
There is no doubt that the body stabilisation works at a certain level and with the obvious benefits of savings having stabilisation across all lenses attached. However it is noted that when shooting with telephoto lenses, the IS / VR becomes much more effective than what the body stabilisation can achieve
Small shift in the lens construct can lead to a large shift in the image projected onto the sensor. Stabilisation in lenses can potentially correct much more than what a sensor stabilised body can do especially on the more telephoto lenses.
If a little angular correction by the gyroscope on the lens can translate to a good several mm of image shift on the sensor, there will be a limit to how much available space the camera can move the sensor around to compensate.
There was a similar discussion here on CS not long ago
"Some of Canons competitors have chosen to use in-body image stabilization. The technique involves moving the image sensor in a controlled fashion, based on signals from movement detecting sensors in the camera body. The obvious advantage of this system is that users have some sort of stabilization available with almost any lens they connect to the body. Short focal length lenses require smaller sensor deflections; 24 or 28 mm lenses might need only 1 mm or so. Longer lenses necessitate much greater movement; 300 mm lenses would have to move the sensor about 5.5 mm (nearly 1/4) to achieve the correction Canon gets with its IS system at the same focal length. This degree of sensor movement is beyond the range of current technology. Short and normal focal length lenses need stabilization much less often than long lenses, so the lenses that need the most help get the least. Further, in cameras with smaller than full-frame, 35 mm film size sensors, equivalent focal lengths become longer, by a factor of 1.5 or 1.6, exacerbating the problem by making all lenses longer."
And a test report of lens stabilisation having an edge over body stabilisation with a few lenses here
I have seen reports demonstrating better outcomes with body stabilisation, but done at short focal lengths with just single test subjects. I believe there will be no arguments on which is better when we test 200mm and above
Longer lenses necessitate much greater movement; 300 mm lenses would have to move the sensor about 5.5 mm (nearly 1/4” to achieve the correction Canon gets with its IS system at the same focal length. This degree of sensor movement is beyond the range of current technology. Short and “normal” focal length lenses need stabilization much less often than long lenses, so the lenses that need the most help get the least. Further, in cameras with smaller than full-frame, 35 mm film size sensors, equivalent focal lengths become longer, by a factor of 1.5 or 1.6, exacerbating the problem by making all lenses longer."
Actually, I think there is a logical error in the above paragraph. I don't think smaller than full frame sensor needs to move a longer distance for stabilisation. Crop factor should not play a part in decided what is the distance to move. In fact smaller sensor should make the system easier to design as the weight of the sensor is smaller, easier to move and accelerate.
Bottomline is in-body IS is much more useful as there are many Canon/Nikon lens without IS/VR like ultra wides and prime lens, ie if you want a stabilized 85mm f1.4 or 12-24mm , in-body is the only way to go
There is no argument to the advantage of usefulness of in body stabilisation. IMO i am happy if the body stabilisation is effective. Does not matter if it is more effective than lens stabilisation.
I am sure there are canon and nikon users hoping that their bodies will come with some kind of stabilisation. After all if it works, i do not mind having my shorter focal length lenses to have some kind of additional stabilisation.
The delay might be less of a marketting issue since lens stabilisation is proven to be really effective, and having a body stabilisation in nikon / canon bodies will be a potential challenge since now the sensor needs to know what the lens has compensated and correct whatever is not corrected.
Actually, many comparison tests have been made between the two types of stabilisation, and in absolute terms, in-lens stabilisation offers better results. However, this is not to say in-camera stabilisation is no good. In fact, it's VERY good (IMHO), just that test results show the in-lens system to be a bit more effective.
But what do you mean by "better"? To me, in-camera stabilisation makes ALL my lenses, past, present and future, effectively stabilised. How many IS/VR lenses can you afford? Is viewing the stabilised effect through the viewfinder important? Not to me, it isn't!
It would probably overcomplicate the design, as the lens and body will need to co-ordinate information to get the best effect. It could also actually cancel out each other's effect and make for a worse overall result.
Don't think of the lens and body as separate entities. For the system to function, it has to work together as a whole unit, so no matter where the point of stabilisation is, the nett effect is still an entirely stabilised system.