New folks keep asking me where to focus, and after starting on ArcView 3.2 in 2001, my advice is to pair timeless fundamentals with skills that follow the long arc — SQL, Python, and working with cloud data — so your maps outlast any single tool. If you’re jumping in now, what’s the part that feels murky, and where can a long-time practitioner help you find the next rung?
Building on @henwhit, start lighter: put one layer in GeoParquet and query it with DuckDB + Spatial (https://duckdb.org) locally, then just swap the path to an S3 URL when you’re ready — same SQL, no server; way faster to grasp than anything in the ArcView 3.2 days. Is the GeoParquet/COG vs shapefile/GeoTIFF split the murky bit, or should I post a 10‑line example?
If the cloud bit feels hazy, try GDAL’s virtual file systems so S3 reads like a local folder: once aws sso login works, gdalinfo /vsis3/your-bucket/path.geojson just opens. From there, point QGIS, rasterio, or ogr2ogr at the same path and you’ve decoupled storage from tools; docs: GDAL Virtual File Systems (compressed, network hosted, etc...): /vsimem, /vsizip, /vsitar, /vsicurl, ... — GDAL documentation. Tiny caveat: on Azure or GCS it’s /vsiaz/ or /vsigs/, and you’ll want to watch request counts so you don’t accidentally DDoS your wallet.
But one thing that’s saved me over and over: before pushing data anywhere, run a geometry sanity check — “ST_IsValid” then fix with ST_MakeValid (or shapely.buffer(0)) — so you don’t chase weird join/render bugs; in ArcGIS Pro, Repair Geometry does the same. @hannah_wilson23 are folks you’re helping getting hung up more on messy geometries or permissions?
I’ve saved hours by choosing the projection on day one — use https://epsg.io to find a local equal-area CRS, “lock the CRS,” and reproject everything to it before measuring or buffering; if you’re targeting web tiles, keep a WGS84 copy alongside… Any region or dataset you’re unsure about for CRS?