Draft mail to CF metadata list

[bekozi]

Discussion relevant to deciding to draft a mail is in https://github.com/bekozi/netCDF-CF-simple-geometry/issues/26. @twhiteaker has offered to start this. The main items to describe and receive feedback on from this mail are:

1. Is our VLEN netCDF-3 approach acceptable?
2. Is a scope of [multi]point|line|polygon + ellipses OK (Ben's approach), or should we design so that we can accommodate all WKT types (WKB-like or WKT approach)?
3. If we plan to support WKT types, do you prefer WKB-like or WKT approach?

[twhiteaker]

Created a wiki page to draft the letter. I'll comment here again when it's ready for review. https://github.com/bekozi/netCDF-CF-simple-geometry/wiki/Letter-to-CF

[twhiteaker]

Letter is ready for review. What do you think of the examples? I copied and pasted from other places in this project to save time, but now I'm thinking all geometry examples should be based on the same geometry. Currently there are different geometries in the simple approach vs the WKB/WKT approach.

Maybe there should be a friendly introduction to the final questions at the end.

I tried to only include the details pertinent to this initial discussion we want to have with CF folks and pointed to our wiki if readers want more details.

[twhiteaker]

I also had a netCDF-4 example and a netCDF-3 example. Maybe I should stick with one or the other.

[dblodgett-usgs]

I've not commented on inclusion of non WKT primitives yet... but I think we should push hard on the core stuff for the standard now with a note that we may want to do parametric stuff later.

I don't think the CF world is going to adopt WKT/WKB directly and the logic that doing the full WKT/WKB in NetCDF without adopting the whole thing is sound.

I'd suggest removing questions for the CF community about the geometry types we'll support. We should say what we've done and why it was worth doing and let people have opinions if they want. Asking for opinions in an open ended way may not go the direction we want.

I do think it would be valuable to ask if people REALLY want more than the WKT primitives, but it should be within the context of saying here's what we ARE doing that's in line with current CF and has somewhat minimal scope.

That's kinda rambling, but do you get what I'm saying?

[twhiteaker]

That's kinda rambling, but do you get what I'm saying?

I think you're saying we show Ben's approach for point, line, polygon, and the multiparts (I lumped multiparts in with the primitives), and ask CF if that's enough. Is that right?

[dblodgett-usgs]

Yeah. I think most of what you've said is right, I'm just suggesting a change of emphasis to say, here's what we have and intend to propose, do people really want more?

[twhiteaker]

Any thoughts on Ben's approach vs. the WKB-like approach?

[dblodgett-usgs]

I thought we had forgone the WKB approach from the reaction to it at the workshop. It's too opaque and non-CF to be acceptable to the community. I missed all the discussion about it though, so maybe opinions have changed?

[twhiteaker]

I really just recall Ethan's reaction about WKT at the workshop and not anyone else's. Actually, if we stick with primitives and multiparts, the WKB-like approach is overkill. I could revise the letter to only include Ben's approach.

Maybe the time to include complex WKT types is when the group working on linked data figures something out, because then you could just link to WKT, GeoJSON, or whatever.

[twhiteaker]

I revised the wiki per comments from @dblodgett-usgs and myself above. You can view prior versions, so to compare the two major versions of the letter:

Just Ben's approach (current version of letter as of this comment): https://github.com/bekozi/netCDF-CF-simple-geometry/wiki/Letter-to-CF/931380687c6ad55aef52eb68d29877346d0d419a

With WKB-Like and WKT approach: https://github.com/bekozi/netCDF-CF-simple-geometry/wiki/Letter-to-CF/6f006d906166bdcaf0e9a6b995758c46486b1d02

[bekozi]

Many thanks @twhiteaker for starting the draft. I thought it covered the basics well without being overwhelming. I reworked it slightly to explain the coordinate indexing approach before jumping into netCDF-3 CRAs. It seems we should focus on geometries and coordinates before jumping into the CRA string example (which is a good one).

The older version includes too much information, in my opinion. If we are able to generate some feedback, we could fish for interest in a more complex WKB encoding and WKT storage directly. In terms of feedback for the WKB encoding method, we likely need to get everyone on the same page with the simpler indexing method. Then blow their minds. I think @BobSimons CRA approach is nice and folks may want to use it for storage outside of any spec.

I think we should make one more sweep of the letter and then make preparations to send it!

[twhiteaker]

Looks pretty good to me. Only thing is maybe we should include the netCDF-3 version of the geometry example at the end of the CRA section.

[bekozi]

:+1: Yeah, that would tie it together.

[twhiteaker]

Done.

[bekozi]

Made a few very minor edits. I think it is ready to send. @twhiteaker, did you want to send to the list?

[twhiteaker]

@bekozi You're the primary contributor of this work, so please go ahead! :)

[bekozi]

Sent. Crossing fingers for useful feedback. :pray:

I'll leave this issue open to discuss any feedback.