SIF end-of-project summary

Download Report

Transcript SIF end-of-project summary

CHS
UCB
CyberCut Retreat, Jan. 2000
SIF -- End of Project Review
Carlo H. Séquin
CHS
What We Have Learned
UCB
Interchange format == design language !
Difference between SIF and SLIDE !

Key issues in design language:
ease of use, high-level expressibility.

Key issues in interchange format:
simplicity, robustness, nonambiguity,
“cleanliness” (SFF: water-tight surfaces).
CHS
UCB
SIF_SFF 1.0 ===> SIF_SFF 2.0
How SIF differs from our initial design:


Shells are connected, water-tight surfaces:

All vertices must be completely shared
by a disk of surrounding triangles.

No complicated (possibly non-planar) faces.

No curved patches (triangle meshes only).

No instantiations of shell fragments
(avoid cracks due to rounding errors).

No triangle strips, fans, arrays.
Floating point numbers are OK:

Receiving end works in FP anyway (unlike EDIF).

The SIF (exp …)-construct was heavily misused.
CHS
UCB
SIF_SFF 1.0 ===> SIF_SFF 2.0 (cont.)


Topology follows from geometry:

Each shell gets topology from explicit edge connectivity.

Boolean operations between closed (positive) shells
allow to specify parts with voids,
and make it easy to design overlapping constellations.
No referencing by names:

Only vertices are referenced for reuse
by array indices within the local scope of their shell.

No reuse of shared surfaces by adjacent lumps
(the exact same vertex coordinates should be
specified again for the surface of the other lump).