Re: New STL format crazy talk.

Date: Tue Jun 02 1998 - 23:35:40 EEST


Change for change's sake is useless. What new RP technology is out there
requiring a new STL format? Why not suggest dumping your computer OS and all
the programs you run because it is in theory possible to make better one? Or
do you happily run the BeOS?

The free market will decide if a "Postscript for 3d" should win out over the
.STL format... and guess what... it did. Pixar introduced the Renderman Spec
in the late eighties as, in their words, "Postscript for 3d." The renderman
spec contains all the 3D geometry data, extensible data forms like nurbs, as
well as color and texture data. Pixar stopped development on Mac and PC
versions of Rman a few years ago, and now only uses a unix version for high-
end filmaking. Why? The software market neither cared nor needed a new

K.I.S.S. The STL format's underlying strength is its simplicity. As computer
memory has grown, and the capabilities of CAD software has evolved to allow
easier, free-form surfaces, STL files have grown bigger and bigger to
accomodate this. The beauty here is that no changes to the spec have been
needed, no software had to be re-written or thrown out, and the STL data that
I made five years ago can be used again today.

Yes, the STL spec is wasteful as it stores redundant vertex information. SO? I
regularly create 100+Mb STL files and easily build them on Actuas at .00125"
Layer Thickness. I have never had problems storing or manipulating such large
files. I have no newer technology requiring color data, higher resolution,
and I have not heard of any new technology that needs something better.

But let's say you've invented something like a new color 3d printer. Do you
need a new data format? Take a look at the STL file spec: for every triangle
there are two unused bytes. That offers the ability to attach a flag for one
of 64K different pieces of information for every tri. Or you could thread
those bytes to store more complex data. (Although a small file, say 4
triangles, might not be able to hold much.)

There is also an 80 byte header that goes pretty much unused.

Sure, 3D's slicing program could use some refreshing. But keep in mind that a
cross section generated through a decently tesselated STL file contains more
than enough info to interpolate a smooth spline curve, with the help of some
AI to find those rare cases when smoothing slices is NOT desired. Also, it
seems 3D has never quite brought the concept of "slice resolution" into the
era of machines that are 30 times faster and have 30 times more memory than
the original SLA slice computers. This explains the zig-zaggy pattern on sla

3D should be prodded at every opportunity to fix these problems, not be
beleagured to waste resources on unecessary formats, so get back in th box,

(In a low hushed voice) We're safe from Al in here.

For more information about the rp-ml, see

This archive was generated by hypermail 2.1.2 : Tue Jun 05 2001 - 22:45:48 EEST