Re: [Geotools-devel] Some thoughts on joining.

[Jody G. <jga...@re...>]

Bryce L Nordgren wrote:

>
>
>If  I may pipe up with a quick use case in the beginning stages of the
>Operations API:
>
>Don't totally discard the push model in favor of the pull model of JAI.  My
>particular application is to process satellite data from the downlink
>station as it comes in.  It would be very handy to set up a JAI-like
>rendered graph through which data could be pushed as it becomes available.
>  
>
Strange I was considering the JAI-like graph as a pull model :-)

In my understanding the  JAI graph does not do anything until you stick 
either a BufferedImage (or a write operations) on the end of the graph. 
I think part of the magic of JAI is that it lets you program in a 
high-level functional way by constructing rendered graph, and then the 
JAI runtime can decide how to implement that graph. It makes fun choices 
for you (like sticking a buffer at any branch in the graph) while 
letting you be explicit (introducing your own BufferedImages into the 
graph for debugging).

I think what I actually admire is this separation.

Cheers