Good First Experiments
Each experiment below comes from a real thread in the dwg7 record, checked on 2026-11-11. You repeat one step of that thread, write down your conditions and your result, and share the record. A finished record counts as success.
Use open data, open tools, and files you may publish. Keep personal data and credentials out of every record.
1. Read one thread from question to result
For a reader of any background.
- Goal: See the whole chain of question, experiment, result, and correction in one thread.
- Inputs: A browser and #1003.
-
Steps:
- Read the first post and state the question in one sentence.
- List the test conditions: hardware, software versions, data, load.
- Find the comment that corrects an earlier claim, and note what changed.
- Visible outcome: A five-line note: question, conditions, result, correction, open item.
- What to record or share: Your five lines, posted as a comment or kept as a file.
- Next: Experiment 9, which repeats a measurement.
2. Overlay two dates of one task
For geospatial practitioners and data contributors.
- Goal: Show a change over time from two public snapshots.
- Inputs: #857 and the public task page it links.
-
Steps:
- Open the task page and note the progress figures with today’s date.
- Return a week later and note them again.
- Place the two dates side by side in a table or a two-colour overlay.
- Visible outcome: One image or table that shows what changed between the two dates.
- What to record or share: The two dates, the source page, and the figures.
- Next: Experiment 3.
3. Observe one block of a mapping task
For geospatial practitioners.
- Goal: Judge the quality of one small block, and say how you judged it.
- Inputs: #857 and the task it links.
-
Steps:
- Choose one small block in a corner of the task area.
- Compare the map with the imagery behind it.
- List three features that match and three that differ.
- Visible outcome: A comment on #857 with your findings for that block.
- What to record or share: The block, the imagery you used, and the criteria you applied.
- Next: Experiment 2.
4. Check a download with a checksum
For anyone who handles large public files.
- Goal: Confirm that two copies of a file are identical before one of them is removed.
- Inputs: Any public file that comes with a published SHA-256 value. #999 shows the practice at 400 GB scale.
-
Steps:
- Download the file and its published checksum.
- Run
sha256sumon the file. - Compare the two values.
- Visible outcome: A match, or a mismatch with a note on where the copies diverge.
- What to record or share: The file name, the size, the checksum, and the date.
- Next: Experiment 5.
- Checked:
sha256sumprinted a 64-character value for a local file on 2026-11-11. The download step awaits a first repeat by a reader.
5. Serve a PMTiles file with range requests
For software engineers.
- Goal: Learn that a single PMTiles file serves tiles through HTTP range requests.
- Inputs: A small public PMTiles file, a static web server that supports range requests, and
curl. -
Steps:
- Serve the file from a folder.
- Request bytes 0–511 with
curl -r 0-511and read the response code. - Open the file in a viewer that reads PMTiles over HTTP.
- Visible outcome: A response code 206 and a map in the viewer.
- What to record or share: The server, its version, and the response headers. #567 records a server switch made for this reason.
- Next: Experiment 6.
- Checked: The static server
http-serveransweredcurl -r 0-511with status 206 andContent-Range: bytes 0-511/2000000on 2026-11-11.
6. Compare a hosted PMTiles file with a local cache
For software engineers.
- Goal: Measure how a local cache changes the load time of tiles from a public host.
- Inputs: A public PMTiles file on a public host, Martin or Caddy, and the repository linked in #642.
-
Steps:
- Load a map from the public host and record the time to first render.
- Start a local cache that points at the same file.
- Load the same map through the cache and record the time again.
- Visible outcome: Two timings for the same view, with the network noted.
- What to record or share: The host, the cache software and version, your network, and both timings.
- Next: Experiment 7.
- Status: The steps follow the source thread or article. A first repeat by a reader, with the conditions recorded, would complete this entry.
7. Repeat a measurement from the Raspberry Pi thread
For software engineers.
- Goal: Add your own hardware to a public comparison.
- Inputs: #1003 and a small computer or a container with a memory limit.
-
Steps:
- Read the setup and the load pattern in the thread.
- Run one server with the same kind of load.
- Record requests per second.
- Visible outcome: One row for your hardware next to the published rows.
- What to record or share: Hardware, versions, data size, load pattern, and results.
- Next: Experiment 1.
- Status: The steps follow the source thread or article. A first repeat by a reader, with the conditions recorded, would complete this entry.
8. Make a PMTiles file from GeoJSON
For GIS practitioners and software engineers.
- Goal: Turn a small vector dataset into one PMTiles file.
- Inputs: A small open GeoJSON file, Docker, and the tutorial Making PMTiles from GeoJSON with unvt (Japanese).
-
Steps:
- Start the container from the tutorial.
- Run tippecanoe on your file with PMTiles output.
- Open the result in a PMTiles viewer.
- Visible outcome: A PMTiles file that displays your features.
- What to record or share: The input source and licence, the command, and the file size.
- Next: Experiment 5.
- Status: The steps follow the source thread or article. A first repeat by a reader, with the conditions recorded, would complete this entry.
9. Style a vector tile map with charites
For GIS practitioners and designers.
- Goal: Edit a map style in a readable YAML file and see the change in a local preview.
- Inputs: Node.js,
npm install -g @unvt/charites, and the article Using unvt/charites for Web Map Styling. -
Steps:
- Run
charites init style.ymlto create a starting style file. - Add one background layer with a colour, then change the colour.
- Run
charites build style.yml style.jsonand check the colour in the JSON. Runcharites serve style.ymland openhttp://localhost:8080/. - Add a vector tile source from experiment 8 and style one layer.
- Run
- Visible outcome: A local preview that shows your colour change.
- What to record or share: The style file, the charites version, and a screenshot before and after.
- Next: Experiment 8.
- Checked: Steps 1–3 ran on charites 2.1.15 with Node.js 22 on 2026-11-11.
10. Build a small elevation tile set
For software engineers and researchers.
- Goal: Encode open elevation data as RGB tiles, and decode one tile to confirm the values.
- Inputs: An open elevation raster for a small region, Python with
rasterio,pip install rio-rgbify, and the article Making small-scale RGB elevation tiles (Japanese). #361 shows the global version. -
Steps:
- Clip the raster to a region of about one degree on each side.
- Run
rio rgbify -b -10000 -i 0.1 --min-z 5 --max-z 9 region.tif region.mbtiles. - Open the MBTiles file with SQLite and count the tiles for each zoom level.
- Decode one tile with the formula
height = -10000 + (R × 65536 + G × 256 + B) × 0.1and compare the range with your input.
- Visible outcome: A tile count per zoom level, and a decoded range that matches the raster.
- What to record or share: The data source and licence, the tool version, the command, and the two ranges.
- Next: Repeat with a second elevation source and compare the two.
- Checked: Steps 2–4 ran on rio-rgbify 0.4.0 with a synthetic one-degree raster on 2026-11-11. A raster that spans the whole globe from pole to pole stops with an error in this version, so a regional clip is the safe start.
11. Summarise a weekly update
For documentation contributors and students.
- Goal: Turn one written update into two clear sentences.
- Inputs: The latest weekly meetup issue, for example #1027.
-
Steps:
- Choose one update.
- Follow its link to the source thread.
- Write two sentences for a reader who joined late.
- Visible outcome: A comment on the meetup issue with your summary and the source link.
- What to record or share: The update you read and the thread it came from.
- Next: Experiment 1.
12. Write a one-page design note
For AI builders and engineers.
- Goal: State a design decision with its context, options, and consequences.
- Inputs: #936, which records a design as an ADR, and one decision from your own work.
-
Steps:
- Read the ADR headings in #936.
- Write the same headings for your own decision.
- Ask one colleague to find the weakest sentence.
- Visible outcome: A one-page note with context, decision, options, and consequences.
- What to record or share: The note, and the change you made after the colleague’s reading.
- Next: Experiment 1, applied to your own thread.