Create README.md

Add the details.

README.md

@@ -0,0 +1,19 @@
+# Issue using xpublish with kerchunk file.
+
+## The issue
+When serving any (an individual time step file or the file that combines the time steps) of the kerchunk files the server throws an error:
+
+```
+ValueError: Encoding chunks do not match inferred chunks
+```
+
+However, when reading the .json kerchunk files directly everything works fine.
+
+## The details
+
+The kerchunk files were made with [this notebook](mk_kerchunk_goa.ipynb) using dachunks = NetCDF3ToZarr(u). I tried various max_chunksize values, but here it is with nothing.
+
+I can read the kerchunk files directly and make a CF data set. [This notebook reads the data and shows a plot](read_json.ipynb) of the second time step to demonstate the reading kerchunk files directly works fine.
+
+I serve the files with xpublish using [this code](server.py). I've tried serving both the merged data set (how I'd like to serve the data) and the data and the grid separately (as shown in the example).
+All attempts (setting the chunks parameter to one time step, to 'auto' and leaveing it off) to serve the data give the same error.