search for: my_nbd

...seems > sensible. > At least unless you realize that having an actual URI it makes sense to > have > that as a part of the scheme, as is customary. > > What I am trying to say is that maybe just the first patch is enough.  When > people want to just write, for example nbd:/tmp/my_nbd.sock by hand, it is > easier than typing the whole nbd+unix:?socket=/tmp/my_nbd.sock.  On the > other > hand, how often do you need to write it manually... My worry is that nbd:/path/to/sock is wrong for use as a Unix socket connection, because it leaves no place for the export name (even...

I'm not sure whether we want to go with just the first patch (reject nbd:unix:/path but still accept nbd:/path), or squash the two in order to go with the second (reject both abbreviated forms, and require scheme://...). Either way, though, nbdkit -U - --run '$nbd' will now error out rather than inadvertently connect over TCP to localhost:10809 instead of the intended Unix connection

...x to show that it is a UNIX socket seems sensible. At least unless you realize that having an actual URI it makes sense to have that as a part of the scheme, as is customary. What I am trying to say is that maybe just the first patch is enough. When people want to just write, for example nbd:/tmp/my_nbd.sock by hand, it is easier than typing the whole nbd+unix:?socket=/tmp/my_nbd.sock. On the other hand, how often do you need to write it manually... Anyway, I don't really have strong feelings about this, so whatever you choose is fine as long as nbd:unix:/path does not try connecting to loca...