Hi Colin, thanks for posting the file and thanks Steve for the analysis.

Yes the edges are just barely out of tolerance from one another to Join.

If you bring it into Rhino, you can use the CrvDeviation command to measure the distance between the edges - on one side it is 0.006 and on the other side 0.008, and they must be within 0.005 to be joined.

If you scale down the model by 1/10 in size (run scale, type 0 <enter> then 0.1 <enter>), then that actually scales down the size of the gaps as well and you can then join it.

On that same file where Join doesn't work, Boolean Union actually doesn't work completely either, it does glue the 2 pieces together but only in one small area and there are a lot of naked edges in there still.


> I thought I'd read that there was some tweaking of
> certain parts to the Join mechanism, so was just
> wondered if this is now because of that??

That most recent change (bug fix) to Join was actually only for curve joining and not surface joining. So I don't think this is from any recent stuff.


Did you use Sweep for these, or Network? I think you may be running into a problem where Network is using a looser tolerance for its internal curve rebuilding, to try and reduce the complexity of the generated surface, but this is leading to sometimes having not quite enough accuracy on the outside edges. I'm going to try and tune that up in v3 to tighten up the tolerance for the outside curves of the network to try and improve that.


But one thing you can try for join problems is to scale down by 1/10 in size, then join, then scale back up again. I probably should figure out some way to have some option for a looser tolerance directly in Join, but since Join doesn't really have any options UI currently and just immediately finishes, it's not so easy to figure out how to make options for it without adding in an extra stage to the command.

Anyway, it seems that Join is ok there, but probably you are running into a tolerance problem in Network.

- Michael