That's not how a chip is scanned for interlace, though. Assuming 1920x1080 chips, then if we consider any single column, you theoretically could indeed read lines 1,3,5 etc to form the odd field, and 2,4,6 etc to form the even - but that's not how it's normally done.
In reality the first field is formed by summing 1+2,3+4, 5+6 etc and the next by 2+3,4+5, 6+7 etc. The two main reasons for doing that are to avoid interline twitter (where a white line on one field, with black lines either side on the opposite field would flicker badly at frame rate) and to improve sensitivity as all the photosites accumulate light all the time. The summation means the signal component adds, the noise component doesn't. Same principle as pixel binning. If it was read as you suggested, each photosite would only be accumulating for half the time.
Taken as a whole, then it follows that for resolution purposes 1+2 will not be the same as 2+3, and neither will that be the same as 3+4. So it should then seem logical to realise that the eventual resolution will be better than 540 - but can't be as good as 1080? It must lie in between the two - yes? I can't remember the maths behind it, but believe the theory predicts 1080/sq rt 2 - about 763 lines. This gives mathematical validity to the observed value of about 0.7 as a Kell Factor - it actually predicts it should be 1/sqrt2 - 0.707.
Originally, Kell factor solely applied to the effect due to video interlacing, but it's meaning has broadened (confusingly, IMO) in the digital world. Originally, it solely referred to how much softening interlace caused - but as that article says, it is now being used to mean how much "real" resolution a digitally sampling display will give relevant to the maximum that may be expected. And no matter whether there is a progressive scan or simply being used for still imaging. Hence, for the same 1920x1080 chip, then if scanned progressively, the max resolution that may be expected is not 1080 lines but 972 if a factor of 0.9 is used. This is why the real resolution of the 1080 system is normally taken to be (even with progressive scan) "about 1000 lpph".
There is a lot more to the topic that could be said, and one factor not so far mentioned is how perceived resolution changes with movement? Practically, it's a good enough approximation to conclude that overall, in terms of resolution, there's not a lot to choose for vertical resolution between 1080i and 720p.
As far as the adoption of HDTV systems about 10-15 years ago, it really needs to thought about more in terms of "1080 or 720" than "progressive or interlace", and there are two very differing cases. For fast motion, the choice at the time was 1080i/25 or 720p/50. For 25fps, it was 1080p/25 or 720p/25.
In the first case, the difference wasn't too clear cut - though generally most people would probably give 720p/50 the edge for broadcast, if only for the compression advantages. In the second case there was no competition - 1080p/25 is vastly superior to 720p/25.
Which is why channels broadcasting mainly sport opted for 720p/50 originally. But because realistically any one channel had to go for 720 *OR* 1080, most channels went the 1080 route if they were showing a mix of 25 and 50fps motion - not because of 1080i v 720p, but because 1080p/25 is so much better than 720p/25.
All agreed. There is little doubt that it's unfortunate 1080p/50(60) wasn't a viable option when the decisions had to be made.
As far as the future goes, then if 8K ever does came around then I hope 24 or 25 fps is totally forgotten about! With the screen sizes that 8K would make any difference for, 24 fps would give such a headache inducing jerkiness as to be virtually unwatchable. I saw the demo footage in London at the time of the 2012 Olympics, and the point was made there that any system really needed a framerate of about 100-120 fps to avoid such. (I think the demo was 60 fps.)
