Background and Goals

[kamicut]

Background

The goal of OSM Tardis is to answer the questions "What if we use cloud native vector formats to parse the minutely replication to prepare stats in the browser?" This would mean the user have to select a bounding box — which then allows us to use things like PMTiles & FlatGeoBuff or GeoParquet.

Goals

This has multiple components:

1. Creating a pipeline of minutely changeset files and turning them into a cloud native format #3
2. Selecting the relevant files on the frontend using a UI
3. Collecting statistics by parsing the files on the frontend and displaying them
4. (stretch goal) Visualizing change over time inside the bounding box

Resources

[internal] https://github.com/developmentseed/labs/issues/330

@wrynearson @batpad @geohacker for adding more detail to this issue

[LanesGood]

@kamicut can we also structure some product goals for this? What would be a workflow, and is there a frontend envisioned for this tool?

Stated another way re: the goal of the project answering the question: "What if we use cloud native vector formats to parse the minutely replication to prepare stats in the browser?"

• who has this question?
• why is the key user asking this question?
• what problem does answering this question solve?
• what is the end-to-end workflow of an ideal scenario of answering this question?

[LanesGood]

Additional relevant comparison stats platforms:

• https://humstats.heigit.org/
• https://osmstats.neis-one.org/
• https://osm-analytics.org/#/

A pitch on OSM Tardis should include why the browser-based stats element is an advantage, how the time element is feasible and advantageous, and how/if a user-stats-focused approach is improved.

[kamicut]

After discussion with @LanesGood, we are attempting to create a targeted frontend and workflow for the tool. It's important to focus on the advantages this tool has over other analytics tools.

The deliverables will include 2 services:

Infrastructure API

The advantage of having a cloud native format and a pipeline for generating these files can be highlighted through documentation on how to query the files over space and time. We should generate examples on how to create a stats collector for a group of users, or even distribute a CLI or SDK for interacting with the API.

Frontend UI

This would have the same capabilities as the CLI:

1. Select an AOI
2. Calculate statistics over a period of time
3. Filter for team of users or specific attributes
4. Display all data added by users over time, maybe visualized using a time slider or some form of recency coloring (color for older features different from color for newer features)
5. "Select a user" from a sidebar or dropdown and filter out all features not belonging to that user

Some stretch goals

• How could we integrate this tool directly with OSM teams to import the teams you're a part of? It would be nice to create a report for a team like "How did HOTOSM contribute to mapping in Beirut in August 2020?"
• What would it look like to have an OSMCha integration? Is it interesting to have stats for an area next to an OSMCha layer to highlight number of edits?
• We can probably do easy counts for features with known tags such as amenities, buildings, highways, etc. How do we do more complicated statistics like the ones in Scoreboard/OSMesa such as KM of roads mapped or KM of waterways added?
• Could we have a user-defined feature filter? For example specifically filter for hospitals or humanitarian structures using the known OSM tag database / iD database?