Saw this blog post earlier today, and now I'm thinking the next step for the node based + publication tool is to be transplanted into a spreadsheet like interface. After all, spreadsheets have proven to be the most successful interface for spatial end-user programming (and i think that is what I'm after).
In this scenario the mechanisms of the [node based] will still stay the same, but instead of connecting wires to reference another node or creating a flow of data, it would be location references. Same as spreadsheets, eg. e4, b8.
When a node is dragged from a position to another, it will also update all its references to its new location.
The components width will be set in -> cols and rows. This component is 3 cols and 4 rows, etc. The values in pixels will be a dependant value on the sum of the rows and cols. The rows + cols can be resized and that would change the pixel size of the components that exist on that axis.
Node connections (lines) can be optional. If references are known, connections can be automatically drawn on demand if connections need to be visualized.
The address of a component will be the top-left cell. A cursor will be implemented that traverses the spreadsheet, when it lands on a components top-left cell, the component is current active component. When a component is active, it receives key presses and can incorporate that to edits its data/ui.