On the names of datasets in GDAL virtual file systems
As you know, GDAL has a virtual filesystem abstraction  wherein datasets are treated as ordinary files at special, private mount points like /vsis3/, /vsihdfs/, /vsiswift/, etc. Rasterio can open datasets using these paths, but favors web-style URIs formatted like "scheme://authority/path".
AWS S3 was the first virtual filesystem supported by Rasterio and we standardized on "s3://bucket/key" for the names of S3 objects. "s3" is not a registered URI scheme, but is used by the AWS CLI and in other projects. Note that AWS more formally uses URNs to identify resources .
Apache Hadoop introduced a "hdfs" scheme  and it seems good to reuse that in Rasterio instead of /vsihdfs/.
"gs" is used for Google Storage in the context of gsutil and BigQuery  and we've followed suit.
As we add support for new storage systems, I'm discovering that some have no concept of a URI scheme. OpenStack Swift is one of these.
I'd like to take the temperature of this group on whether to coin a URI scheme for OpenStack Swift or to punt for now, rely on /vsiswift/ filenames, and see if the Swift project and its users rally around a particular URI scheme. Thoughts?