Thanks for your feedback, Joojaa.
I wholeheartedly agree for the most part. You raise some very valid points.
I do think, though, that in some cases, lets say I need the vertices for baking out some position locators for export to Nuke or whatever, EVEN in a large production this will yield much better results (as a one or five time use thing).
I could now, after a bit of scripting, have used it for what I needed and published the position locators to whoever requested them (or do something else with them, bake whatever result, it being particles, influences geometry or whatever).
My point is, most people aren't technical artists or TDs and often, if you know a bit of mel (or in most cases even better; Python) you can get things done without passing it onto the TDs queue of things to do, waiting for it to be ready and then go back the shot (because you had to do something else while waiting) bla bla bla..
If you want to get something done.....
However, I too think that particle instancing would probably be a much better way to do something like this.
I disagree with a couple of points, though:
ScriptJobs DO work. In this case we needed the scriptJobs to update the locators when changing the subdivision. I'm not going to change that during dg evaluation anyway, so that point isn't really relevant.
Which nodes would you rename in order to break it?
The scriptNode preserves the connections to the central nodes, so tahat shouldn't be a problem.
You are correct about deleting or renaming the plane, but thats easy to fix. Just add a dummy float attr to the plane and have the expression do something with it.
Most problems like that shouldn't matter (and if you're DELETING the plane it probably means you don't need the locators either, so get rid of it all).
Yes it isn't fool proof and not very flexible, but when you have a task at hand, you usually try and do the least amount of work (unless you are actually writing this as a general script and plan on releasing it). Doing otherwise would be like a modeler modeling the whole house, no wait, the whole street in case we need to bring the camera out of the living room (where the shot is).
I'm speaking as an artist who use mel to get things done and also set up shot- or project specific tools and not as a software/pipeline engineer.
I DO get what you are saying (and you also point out that "if a hack is enough..", and that's what I think might be enough in this case (depending on the further usage of the setup).
By the way, what did you mean about ONE namespace and ONE namespace only?