On Fri, 24 Oct 2003, marius schebella wrote:
I don't know jitter, but maybe pdp goes a bit into that direction? plz correct me!
I don't know what makes you believe that.
As far as Tom told me, PDP operates using a system of types. For each named type of data there is a way packets are constructed and a set of object classes willing to accept them.
GridFlow is less typed, as in, there is basically one data type, the Grid, and then it is parametrized by a list of nonnegative integers representing its size in each dimension, and a scalar number type (at the machine level, specifying bits-per-value and such: uint8 is 8 bits).
For example, Grid[240,320,3,uint8] would be a type in GridFlow (not that I actually use that syntax in the system itself). Usually, something of a size 240*320*3 could be assumed to be a RGB picture by the user, but technically, it can be anything, and most object classes actually don't give a damn. The user is supposed to know, and most users would only use RGB pictures most of the time, so they don't really have to ask themselves that question. But a Grid can also be used to represent other types of data, like:
- a polygon
- an indexmap (for morphing pictures)
- a colormap (palette)
- a whole series of pictures (a short movie loop or an icon collection)
- a dimension-list (representing the size of another grid)
- a single number (0-dimension grid: a scalar)
- emptiness (0-element grid)
- convolution matrix
- linear transform (change-of-axis) matrix
- and so on
And then some objects can work and are meaningful on all those kinds of data, even though those objects may not know what they are working on.
This is what GridFlow introduced as a new way of doing things back in the spring of 2001, and what Jitter reproduced [albeit with a bigger budget and better leverage] about a year later.
AFAIK, PDP has linear algebra functions, and actually is better at it than GridFlow, but those functions are separate from the picture processing, so for example in PDP you can't implement your own hue-shift effect by multiplying a picture by a matrix, whereas I'm doing exactly that in GridFlow. But maybe Tom can give you the exact details, as I am but a beginner with PDP.
Note: I don't know what NATO allows, and whether historically it's in the same streak as GridFlow. I never got to try, I never got to see it running, the company has disappeared, and its assumed author won't explain her design in a honest attempt to communicate. No matter what actually was available in the mysterious NATO, all my inspiration for GridFlow came from a 40-year old programming language called APL.
Mathieu Bouchard http://artengine.ca/matju