probably sooner, I think Sony has already switched ram from XDR to XRD2. The RSX is getting shrank pretty quick by nVidia. Should be 40nm by the end of the year. CELL will be 32nm by the end of the year or beginning of next.

This could also mean that PS3 will continue it's course
and they will develop cheaper, developer friendly, and casual player friendly console.

Developers would be more practiced at it, for sure. And it would be good that they aren't making it even harder to program for, but it certainly wouldn't be 'easier' by any sense

 but they would of have to make a slimline ps3, the present one looks too big and scary for little kids .

Few more cores and SPEs wouldnt effect anything since the base coding knowledge would be there.

Actually, it would be harder because now they have to find a way to take linear algorithms and make it scale to 20+ SPEs instead of just 7. However, if IBM makes the improvements to the Cell that they spoke about, a few technical challenges will become much easier to work around. I guess in the end it would break even to be as difficult as it is now.

While I'm not going to make any specific comments about the rumour, what Sony is working on today and might get leaked as a rumour may not be much like what Sony settles on a couple of years from now when they finalize the hardware design for the PS4.

Actually, the specs being leaked and predicted by people with technical knowledge is actually on par with the technology that will be available in 2011 or so. A cell with 24-32 SPEs is very much plausible.

It really depends on the direction they choose to go with on a "Cell V2" ...

Hypothetically speaking, Sony could choose to produce a processor with 3 to 8 Power Processing Elements and keep the same 8 SPEs just for backwards compatibility, and the end result would be (somewhat) similar to most other multi-core processors and could be much easier to develop for.

The problem with that design is that the SPEs would be neglected since the PPEs are much easier to develop for and it would be like programming 2 architectures because the SPEs and PPEs have a few differences. I think that would actually complicate things a lot. The job of the PPE is to manage and distribute tasks to the SPEs and handle a few tasks which requires data from all SPEs. An equal number of PPEs and SPEs would break that philosphy and require major changes to their existing codebase.