it use a tree to store the filesystem instead of a flat file list (for better perfs, similiar to InMemoryFileSystemProvider)
it implement the file stream api
it doesn't rely anymore on the file content to get the file size => removing the need to get the file twice (one from the stat to get the size then another one to get the content) => removing the need to memoize the file content => stop keeping all the file content in memory, reducing memory footprint
What it improves:
stat
to get the size then another one to get the content) => removing the need tomemoize
the file content => stop keeping all the file content in memory, reducing memory footprint