> If an http URL is pointing to a file that is accessible via
> NFS (i.e. the file resides in a directory mounted by the client),
> then in principle the client can read the file directly from the file
> system rather than go via HTTP to the server, essentially
> treating the URL as if it was of type "file:" rather than "http:".
Yes, my browser works that way. It does improve performance, especially
when the server would otherwise have to use nfs itself to access the file
in addition to sending the file via HTTP.
The browser is left to interpret the file type on the basis of file name
suffix or pattern match on data contents. I have this code left over from
the old HTTP/0.9 days when you had to guess the file type.
Dave Raggett