Read the endpoint fields

Priority chooses a tier, weight orders peers within a tier, and port names the destination endpoint.

• The textual owner/TTL/class/type precede SRV RDATA. RDATA itself is Priority(16 bits), Weight(16), Port(16), then Target encoded as a DNS name (format, pages 3–4). All three numbers range from 0 through 65535 and use network byte order, most significant byte first.
• Priority is a preference tier: a client MUST attempt the lowest-numbered priority it can reach. Within one priority, hosts SHOULD be tried in an order defined by Weight. Weight is not a way to override a lower-numbered priority.
• Weight is a relative selection input among entries with the same priority. Larger weights SHOULD have proportionately higher selection probability. Administrators SHOULD use zero when there is no selection to do. Port is the service’s actual port on that target, which need not be the conventional assigned port.
• Chosen RDATA: priority 10, weight 3, port 9, target a.example.com. The first bytes are 00 0A 00 03 00 09. The uncompressed target name uses label lengths 01,07,03, then a root zero, totaling 15 octets; RDATA totals 6+15=21 octets. This is a fictional service endpoint, not a claim that TCP port 9 is a registered foobar service.
• Target MUST have one or more address records and MUST NOT be an alias. RFC 2782 says name compression is not to be used in Target unless permitted by future standards action (page 4). This guide encodes the uncompressed case; it does not silently replace the name with a DNS compression pointer.
• The geometry here scales to RDATA octet counts: three 2-octet fields followed by a 15-octet name. The name’s displayed dotted form is a reading aid, not its wire encoding or a guarantee of server identity.