search

ogcscotts / TC-Meeting-topics

place to discuss topics raised by Working Groups
10 stars 0 forks source link

Requirements for a Vector Tiles standard #55

Open jyutzler opened 6 years ago

jyutzler commented 6 years ago

OGC has conducted more than one TC ad hoc session regarding vector tiles, but this has not led to a move to charter a working group or even any kind of consensus on what sort of activity this working group would attempt to accomplish.

jyutzler commented 6 years ago

I believe that we have been overthinking this issue. Let's establish a 2D standard and declare victory. While 2D/3D/ND is a valid requirement, there is no benefit to conflating them into the same standard. There is no clear need for 2D presentations to support 3D vector tiles and vice versa so why bother? OGC has already adopted I3S as a community standard and is considering 3DTiles as well but a 2D approach remains unresolved. Let's fix this.

To get there we first need to nail down the conceptual model for tile matrix sets. Fortunately the announcement of the open comment period for the draft standard for tile matrix sets may be made by the time you read this. Once that is in place, I propose to use it as a basis for a 2D vector tile solution, with multiple encodings being possible.

jerstlouis commented 6 years ago

Jeff, at the first ad-hoc we tried to establish a working group and started putting a charter together. Then there was some questions about the organization of working groups at the OGC at the plenary and no one was willing to advance the idea of a new working group for tiling anymore.

So we decided that Architecture DWG and OWS Common SWG could suit the purpose of advancing these vector tiles standards, and at the second ad-hoc we tried to focus on setting up an interoperability plugfest, to show that we do have enough to write down standards (I think we at least want to cover extensions for WMTS, and extensions to WFS).

I agree that the tile matrix set is something we want to leverage. (The document is available at https://portal.opengeospatial.org/files/?artifact_id=79767&version=1 ) Also agreed multiple encoding is a must!

As for 2D vs. 3D vs. ND contents, conceptually, if you already have a tiling scheme partitioning the world in geographic regions, you can re-use these 2 dimensions (e.g. latitude & longitude) and add additional ones such as time or altitude as either additional tile "secondary columns", as parameters, or being embedded in the payload of the tile itself. So I would argue that the same tile matrices can easily be re-used together with a time dimension, can be extended to cover different altitudes and/or can contain 3D contents. But I agree that keeping all this into consideration should not preclude making progress on the urgent needs of establishing a standard for 2D vector tiles.

See also discussions in https://github.com/opengeospatial/ideas/issues/62 and https://github.com/opengeospatial/ideas/issues/58

PeterParslow commented 5 years ago

On reading the draft "Features and their geometries", I think this 'abstracted' discussion of partitioning - so it isn't "just" 2D spatial - has relevance to the classification of the world into 'feature instances'.

My thesis: we have a tradition of assuming that feature instances are spatially disjoint; in some cases ("skin of the earth") they specifically form a coverage (no gaps, no overlaps); in others, we allow gaps & overlaps - but a given 'bit of 2D/3D space is primarily considered as being 'in' one feature instance'.

In the 'new world' described in 'features & their geometries', this isn't necessarily the case: feature instances can 'claim' multiple feature types. My concern - and the overlap with here - is that in order to partition the world into instances (or tiles), we need to choose one (or a small number) of parameters on which to partitiion - geometry, ownership / administration, reflectivity to particular light/radio waves,..., probability of rainfall in a particular period. Those are examples from the 'feature' world. Hopefully I'm not the only one who can see the similarity with this tiling discussion?

jyutzler commented 5 years ago

Given that the Vector Tiles Pilot is well under way, can this issue be closed?

cnreediii commented 5 years ago

Actually, the first step is defining and agreeing to an Abstract Spec for tiling. A draft now exists. The conceptual model is very high level and apropos to any dimension. There is currently only one logical model for the 2d Euclidean case (sorry Peter) but other logical models for other dimensions can easily be added.