Thanks for your reply. I am using the same version, Version: 2.88.1144.0 64-bit (desember 2020). The URL is correct, I enter the same URL to download it to local and I see also the preview.
It seems that it depends on the size, or a particular value that is being imported. When I have a parquet file with 10000 rows/148 KB, 50000 rows/648 KB, 100000 rows/1.3MB, it works fine, but 500000 rows/6.2MB fails when using a direct link. I get to see the truncated preview, but when loading it gives the error message: "Parquet.Document cannot be used with streamed binary values."
However, when I download the same file loading works fine.
As an additional test, I have concatenated 5 times the 100000 same rows that worked fine, and this gives the same error, ruling out that it is a particular value in the parquet file that produces this error.
I suspected it had something to do with the SNAPPY compression because it is not splittable, but the same parquet file using no compression causes the same issue.
To reproduce, I have a file hosted at http://www.datec.no/dev/public.parquet that is based on the public available data.gov dataset (Allegheny County Property Sale Transactions).