If the developer is going vector, they don't really have to do anything (the scaling works transparently). And its turning out to be a good time to go vector.
This is quite true, with existing technology the graphic designer always has to touch up their Illustrator PNGs in Photoshop (conspiracy theory: does Adobe do this on purpose so people will buy photoshop?). But at higher PPIs, the aliasing effects of rasterization are much less, you could almost turn anti-aliasing off at retina display levels, so vectors begin to make more sense.
Of course, the problem is that there are still a lot of displays out there where anti-aliasing is necessary, and for some reasons our algorithms always screw it up if done in real time. Perhaps we just need better algorithms (e.g., see Direct2D).
The final problem they mention is one of physical size, but it makes sense that you'd have a lesser detailed vector for something that is really small vs. something that is very big; its a design issue.
Comments
The TL;DR version:
The developer is responsible for a lot of the screen dimension differences using flexible layout and progressive enhancement/degradation.
"Retina" scaling is fixed at 1.0, 1.4, and 1.8 (100%, 140%, 180%).
If the developer is going vector, they don't really have to do anything (the scaling works transparently). And its turning out to be a good time to go vector.
There was an article on HN recently that demonstrated why vectoring is completely unviable for this: http://www.pushing-pixels.org/2011/11/04/about-those-vector-...
In short, vector graphics scaled up simply can't compare to graphics that were hand-made for a specific size, as the article shows.
This is quite true, with existing technology the graphic designer always has to touch up their Illustrator PNGs in Photoshop (conspiracy theory: does Adobe do this on purpose so people will buy photoshop?). But at higher PPIs, the aliasing effects of rasterization are much less, you could almost turn anti-aliasing off at retina display levels, so vectors begin to make more sense.
Of course, the problem is that there are still a lot of displays out there where anti-aliasing is necessary, and for some reasons our algorithms always screw it up if done in real time. Perhaps we just need better algorithms (e.g., see Direct2D).
The final problem they mention is one of physical size, but it makes sense that you'd have a lesser detailed vector for something that is really small vs. something that is very big; its a design issue.