Notes for TSC meeting.

Signed-off-by: Jeff Lait <jlait@andorra.sidefx.com>

tsc/meetings/2023-09-19.md

@@ -0,0 +1,97 @@
+Minutes from OpenVDB TSC meeting, September 29th, 2023
+
+Attendees: *Ken* M., *Jeff* L., *Andre* P, *Dan* B.
+
+Regrets:
+Attendees: *Greg* H.,  *Rich* J. *Nick* A.
+
+Additional Attendees: 
+Karl Marrett
+
+Agenda:
+
+1) Confirm quorum
+2) Secretary
+3) Explicit Instantiation of the Tree
+4) VTT Integration
+5) Reproducible Workflows
+6) Next meeting
+
+------------
+
+1) Confirm quorum
+
+Quorum is present.
+
+2) Secretary
+
+Secretary is Jeff Lait
+
+3) Explicit Instantiation of the Tree
+
+We are still implicitly instantiating the Tree.  But we hadn't got
+that far in the explicit instantiation.  If Tree was instantiated, the
+function templates inside get instantiated, and it becomes a big
+rabbit hole to get it working.  Also the goal is to remove stuff from
+the Tree class to make this faster.
+
+Tools were the first layer to be done.
+
+4) VTT Integration
+
+Cookbook would suffice rather than actual code for establishing.  Do
+we discuss what the definition of Tile is?  How do we make sure our
+terminology is consistent?
+
+What of naming?  VTT is confusing with VDB.  What about Multires or
+Adaptive grid?  Vs Sparse grid?
+
+Can we make Grid the new baseclass?  It has a Tree() that returns a
+reference so is hard to nullify.  
+
+But TreeBase could the right point to inject?
+
+There is no activation in VTT.  But there is a box from the 2^32
+range.  An ROI mask could be done with a VDB or a VTT with coarse
+tiles.
+
+What does it look like to convert between them?  Does VDB become the
+finest resolution of the VTT?
+
+By Friday we'll share the google doc.
+
+VTT should be a first class citizen.
+
+5) Reproducible Workflows
+
+Karl Marrett presented a proposal to improve OpenVDBs
+interoperatbility with package systems, nix in particular.
+
+Add some files to root of repository to make it understandable by
+package managers.
+
+Compliment of openvdb is distribution of VDB.  So if we can
+commoditize distribution, we'll improve things.  Package managers are
+not commoditized, they are slow, specific, and often useless for
+active developers.  
+
+Git sort of transcends package managers, it has the idea of forking,
+etc.  Nix Flake is a decentralized package manager.  
+
+Takes in flake.lock - specifies specific revisions of different repos.
+Can output to a docker-like file.
+
+Semantic versioning is built-in as well.  So can auto-bump based on
+latest versions.
+
+New libraries would have to be added in if they were not already in
+NIX.
+
+NanoVDB is not currently supported.  MacOS is not supported as there
+is a broken package somewhere.  
+
+Can we get a CI job on it?
+
+6) Next meeting
+
+Next meeting is on September 26th, 2023. 2pm-3pm EDT.