Yes? No?

I have had a couple friends try to talk me into playing that.

See:

http://www.cemetech.net/forum/viewtopic.php?t=5025

No, you should be staying loyal to Freebuild instead! And elfprince needs to get on Freebuild so we can tap some of this sandbox world enthusiasm.

Sorry, Kerm.
Freebuild fails on me whenever I try to customize.

Have you tried good old TBM2, then? Sorry to hear about the crashes; you should tell Elfprince or CyberPrime so they can implement fixes.

*can* implement fixes.....

Elfprince is making progress at last; I have faith in him.

fastforward 1 week and it's back to life.

I played Minecraft a few months back, it was atrociously buggy, and I lost my Firefox session with >30 tabs, because it ate my input devices and wouldn't give them back whenever it loaded, so I had to force a new session.

He has indeed been making lots of progress.

Anyway, Minecraft reminds me of the Cube FPS engine. Looks interesting, but I think I'll pass for now :B

EDIT: elfprince13, gehy!
I guess it's probably a good idea that I pass on it then >.<"

My friend Adam posted this to my facebook. I had to share

Cool stuff, and good for him for coming up with that kind of scripting. Makes me ponder how similar things could be done in TBM/Freebuild...

Minecraft (and Blockland) definitely sacrifice brick detail for rendering speed. Lego Universe makes use of cloud rendering. My hope is that if I implement proper geometry instancing for studs and CHC++ culling routines, most builds will scale pretty will and we'll be able to get away with the higher complexity brick models, by only drawing the more complex parts in places where they are actually visible. If I really have to I'll end up implementing primitive substitution, as was my original plan, either as a full fledged replacement for studs, or as part of LOD scheme. That's further down the line though.

Yeah, I think that's very important. Heck, if it came right down to it, I'd be willing to lose studs entirely or go with the old Blockland model of printed-on studs for the sake of build complexity vs. framerate.

The nice thing about the studs is that because they rest on top of the bricks (and don't leave holes), I should be able to fairly easily scale the detail from none at all, to decals, low-poly studs (maybe rectangular), the usual ones, or even the super high quality ones.


With regards to the train system, I think it should be relatively easy to store the track as a graph, where edges are splines and vertices (at intersections/rail switches/and connecting rail segments) have a function that maps incoming edges to outgoing edges. Saving it could be something as simple as writing a list of indexed vertices labeled by type and location, and a list of edges defined by pairs of vertices and the equation of the curve to trace between them. All lego tracks curve only in 2 of the 3 spatial coordinates, so the calculations for a Frenet-frame based follow track constraint in Bullet should be trivial to implement, and really just have to be an extension of the slider constraint where the axis to follow is not a straight line, but recomputed at each tick based on the tangent of the curve to be followed. We will ALWAYS have the plane normal or binormal based on bounding box calculations and track orientation, and if we use polynomial or trigonometric splines the tangent should be easy. With two out of three, we just need a quick cross product for the third. I've written functionally equivalent code in Maple last year, so it would be a matter of porting it to our Vector library instead of the Maple VectorCalculus package.

Elfprince, that sounds great! And it occurred to me that if one end of a train continues off of a graph, it can just become a normal kinematic object until it's placed back on a track. I suppose we'll have to have a small amount of tolerance for the purposes of re-railing runaway trains.

Right. Though for placing track-vehicles, and for placing the tracks themselves, a mode that only lets you place on, and at the end points of, existing track would be easier. We could say that if a track-vehicle collides with a loose track end, and is oriented/traveling a long a vector within a pi/4 cone of the tangent vector we will re-rail it.


On a side note, more related to Minecraft, this video of an ALU just made it on slashdot.

Wow, and people have accused me of having a nerdy voice in some of my voiceovers. This guy blows me out of the water. Pretty impressive, I guess? Nothing we couldn't do in FB, though.

Well, I think the impressive part is that it is functioning.