| ▲ | gruez 5 hours ago | |
>Barely enough time for the domain to resolve! contrary to popular belief, DNS doesn't actually need 24 hours (or whatever) to "propagate". It only needs enough time for caches to expire, if they exist. For non-existent domains this shouldn't be an issue. | ||
| ▲ | gnabgib 5 hours ago | parent | next [-] | |
DNS controls the length of time an answer is valid (with time-to-live/TTL), some (poorly configured, but for a historical good reason) caching servers ignore the TTL and keep the record much longer (~24 hours). A non-existent (not yet resolvable) domain, will get an NXDOMAIN response which also gets cached via a TTL setting (which again might be ignored by caching servers). All that caching - while the current response is still "valid" because the TTL hasn't expired - is what people call "propagation". | ||
| ▲ | adrianmonk 2 hours ago | parent | prev [-] | |
Aside from regular caches (that use TTLs), there can also be a replication delay between primary master and secondary master servers. You make your changes to the zone file on the primary server. The secondary replicates the zone file (a "zone transfer") and uses its copy to serve requests, enabling you to have multiple authoritative servers. In the old days, the secondary server simply polled the primary server at a fixed interval to see if its zone data was out of date. It could potentially be hours if that's the schedule you chose. It wasn't uncommon to restart the secondary servers by hand if you needed something to propagate faster. Later, more sophisticated stuff was added like incremental zone transfers and the ability for the primary to send a notification to the secondary instead of relying only on polling. Those made it more practical to replicate quickly. There were scenarios back in the day when quick replication wasn't worth the cost. The primary and secondary servers might be on opposite sides of a 9600 baud modem connection where zone transfers would eat up a significant portion of available bandwidth. TLDR: DNS "propagation" is real, not a myth. It's just that today it's pretty trivial to make the propagation delay so tiny that it's negligible. | ||