FeatureServer vs MapServer: which one gives you the data?

The short answer: a FeatureServer serves features. A MapServer serves pictures of features, and sometimes the features too. An ImageServer serves rasters. If you want data, look for FeatureServer first.

Knowing which one you are looking at, before you start, saves the hour spent trying to extract geometry from something that only has pixels.

The three you will meet

FeatureServer

Serves the actual features: geometry and attributes. Built for exactly what you want.

• Its /query endpoint returns GeoJSON or Esri JSON.
• Each layer describes its own fields, types, and extent.
• It may support editing, if the publisher enabled it. You will not have permission.

If a public web map is backed by a FeatureServer, you can almost always get the data.

MapServer

Serves rendered images. You give it a bounding box and a size; it gives you a PNG with the data drawn on it. The cartography happens on the server.

The catch, and it is the useful one: a MapServer often also supports Query. Many are published from the same source data with query enabled, in which case the same /query?where=1=1&outFields=*&f=geojson trick works exactly as it does on a FeatureServer.

So do not write off a MapServer. Check it.

ImageServer

Serves raster data: imagery, elevation, and the like. There are no vector features to extract. You can get pixels, sometimes with real values behind them (an elevation ImageServer will tell you heights), but there are no polygons in there.

How to tell in ten seconds

Paste the service URL into a browser. ArcGIS REST endpoints describe themselves in readable HTML.

Look for Supported Operations near the bottom:

• Query present: you can extract features. This is the one that matters.
• Export Map but no Query: images only.
• Export Image, Identify: an ImageServer.

Then look at Max Record Count, usually 1000 or 2000. That is how many features you get per request, and it tells you whether you need to page.

If you would rather not check by hand, Geo Hound labels each service with its exact type as it detects it, and reads the same description automatically, so a card tells you what you are dealing with before you click anything.

The decision, in practice

| You see | Do this | | --- | --- | | FeatureServer | Query it, or add it to QGIS. You are done | | MapServer with Query | Same as above. Treat it as a FeatureServer | | MapServer without Query | Images only. Go looking for the source data elsewhere | | ImageServer | Raster. Fine as a layer, no features to extract | | VectorTileServer | Geometry, but tile-clipped and simplified. See finding tile URLs |

When there is no Query

You are not stuck yet. Try these, in order:

1. Drop back to /rest/services. Many servers list their whole directory. The same data is often published a second time as a FeatureServer for exactly this reason.
2. Look for an open data portal. Search the organisation's site for the layer name. Councils and agencies routinely publish downloads alongside their viewers, and the download is better than anything you would extract.
3. Check for a WFS. Some organisations run both, and a WFS gives you features. Finding the WMS URL applies: WMS and WFS often live on the same server.
4. Ask. Genuinely. Email whoever publishes it. Public sector GIS teams are frequently delighted that somebody wants their data and will hand you a file. Nobody tries this, and it works more often than it has any right to.

Why the whole family matters

Esri publishes seven REST service types: MapServer, FeatureServer, ImageServer, GPServer, GeometryServer, GeocodeServer, and VectorTileServer. Three of them hold data you might want; the rest are machinery.

Geo Hound detects all seven, alongside the OGC and OGC API families, so what is behind a viewer is named for you rather than guessed at. Then it hands you the URL formatted for QGIS or ArcGIS Pro.