This is an interesting project done at MIT using OpenSCAD, where they
created a "customizer" interface that they claim is far superior to what
Thingiverse has.

press release: http://news.mit.edu/2015/customizing-3-d-printing-0903

tech report: http://cfg.mit.edu/sites/cfg.mit.edu/files/fabforms_paper.pdf

video: https://www.youtube.com/watch?v=LVOVmIIbeTY

One thing missing from OpenSCAD, needed by this project, is the ability to
test a combination of parameter values and report an error. There is a
recent thread, "Quit or Exit function?", that discusses this deficit.

As a general approach, it looks interesting. However, in typical academic
fashion, it lacks practicality. They precompute data sets to make later
operations faster. That is a pretty standard CS technique, but the simplest
model took 1 day of CPU time, and the most complex 22 days, and also
produced up to 15GB or so of cached data.

So while they criticise Thingiverse, Thingiverse probably have thousands of
models. Precomputing the data and storing large amounts of data would be a
significant overhead. Researchers might get their 40 Amazon cores cheap or
free, but others have to pay.

So yeah, if you throw CPU and memory at a problem you can do it quicker.
What would be novel is to find a way to do it without incurring a lot of
overhead.

Obviously the press release is a puff piece, but it doesn't mention all the
third-party work they built their stuff on.

--
View this message in context: http://forum.openscad.org/MIT-FabForms-research-project-tp13694p13697.html
Sent from the OpenSCAD mailing list archive at Nabble.com.