The
dns-sd command is a network diagnostic tool, much like
ping(8) or
traceroute(8). However, unlike those tools, most of its functionality is not implemented in the
dns-sd executable itself, but in library code that is available to any application. The library API that
dns-sd uses is documented in
/usr/include/dns_sd.h.
The
dns-sd command is primarily intended for interactive use. Because its command-line arguments and output format are subject to change, invoking it from a shell script will generally be fragile. Additionally, the asynchronous nature of DNS Service Discovery does not lend itself easily to script-oriented programming. For example, calls like "browse" never complete; the action of performing a "browse" sets in motion machinery to notify the client whenever instances of that service type appear or disappear from the network. These notifications continue to be delivered indefinitely, for minutes, hours, or even days, as services come and go, until the client explicitly terminates the call. This style of asynchronous interaction works best with applications that are either multi-threaded, or use a main event-handling loop to receive keystrokes, network data, and other asynchronous event notifications as they happen.
If you wish to perform DNS Service Discovery operations from a scripting language, then the best way to do this is not to execute the
dns-sd command and then attempt to decipher the textual output, but instead to directly call the DNS-SD APIs using a binding for your chosen language.
For example, if you are programming in Ruby, then you can directly call DNS-SD APIs using the dnssd package documented at
<http://rubyforge.org/projects/dnssd/>.
Similar bindings for other languages are also in development.
dns-sd -R name type domain port [key=value ...]
register (advertise) a service in the specified
domain with the given
name and
type as listening (on the current machine) on
port.
name can be arbitrary unicode text, containing any legal unicode characters (including dots, spaces, slashes, colons, etc. without restriction), up to 63 UTF-8 bytes long.
type must be of the form "_app-proto._tcp" or "_app-proto._udp", where "app-proto" is an application protocol name registered at
http://www.dns-sd.org/ServiceTypes.html.
domain is the domain in which to register the service. In current implementations, only the local multicast domain "local" is supported. In the future, registering will be supported in any arbitrary domain that has a working DNS Update server [RFC 2136]. The
domain "." is a synonym for "pick a sensible default" which today means "local".
port is a number from 0 to 65535, and is the TCP or UDP port number upon which the service is listening.
Additional attributes of the service may optionally be described by key/value pairs, which are stored in the advertised service's DNS TXT record. Allowable keys and values are listed with the service registration at
http://www.dns-sd.org/ServiceTypes.html.
dns-sd -B type domain
browse for instances of service
type in
domain.
For valid
types see
http://www.dns-sd.org/ServiceTypes.html as described above. Omitting the
domain or using "." means "pick a sensible default."
dns-sd -L name type domain
look up and display the information necessary to contact and use the named service: the hostname of the machine where that service is available, the port number on which the service is listening, and (if present) TXT record attributes describing properties of the service.
Note that in a typical application, browsing happens rarely, while lookup (or "resolving") happens every time the service is used. For example, a user browses the network to pick a default printer fairly rarely, but once a default printer has been picked, that named service is resolved to its current IP address and port number every time the user presses Cmd-P to print.