I understand there are multiple lifecycle statuses that a domain goes through before it's effectively "dropped", and these may vary per tld. (https://www.icann.org/resources/pages/gtld-lifecycle-2012-02-25-en) But once it is available, that very nanosecond, what are the factors involved in thr domain being won by registrar A and not B?
what are the factors involved in thr domain being won by registrar A and not B?
Luck. Or sheer force. Or many other factors that no one will discuss publicly, for obvious reasons (multi-million businesses have been built just on top of that, so...)
From your comment:
Are they for example just internet speed? And then physical distance? Distance from what? The registrar servers? ICANN? If multiple requests are handled by the icann registration server, can it prioritize a response over the other?
- ICANN has no play in the registration path or any other part of a domain name lifecycle, so that can be taken out of the equation
- registrars connect to registries to conduct operations; so on a pure theoretical level, the closer "network wise" the registrar emitting server (client side of the EPP connection typically) is from the receiving server (server side of the EPP connection), the better chances it has; for this reason some registrars can try to find out how a registry is connected, typically which Internet exchange points it uses, and try to be there
- so obviously "Internet speed" does play a part, but honestly I wouldn't consider it to be the most important factor, except in very specific high impact cases; it is not like the HFT sector where nanoseconds can really be millions of dollars in consequences
- rephrasing your last part "If multiple requests are handled by the registry, can it prioritize a response over the other?", technically yes, obviously, contractually, no, but this is impossible to prove. Almost, if not all, registry contracts will outline very clearly that all registrars are on a level playing field and none is favored over others. This is also an ICANN requirement towards all gTLDs.
- but then things are more complicated, mostly because various registries decided to put in place specific systems to handle "drop catching". It can go from a penalty system (if a registrar tries to register a domain just about to be released and failed it got a penalty point and after some level of penalties it is barred to do any other operations for like 24 hours; so here obviously a registrar trying to catch only one name will have more chances globally than a registrar trying to catch 100 names on that day, as penalty points may accumulate quicker), to a specific server to use, which may need specific contracts and payments, and registrar can pay for more connections (so here technically a registrar with biggest wallet has more chances, which is also why 20 years ago hundreds of registrars were created just to play in this game)
- so the above point if very important regarding amount of names (on a given day) that a registrar will attempt to catch: a registrar seeking only one may have more chance just because all of its resources will be targeting that specific domain, where a registrar with multiple requests will have to share its resources among all names, so let us say it has 10 outbound connections but 100 domains to handle, so obviously some will not be in the first 10 batch, no matter what, and if another registrar has a single name and send it immediately, then the other registrar loses
- in some cases some registrars will go from great length (and I witnessed that first hand working inside a registry) to try to reverse engineer the algorithm that decides when the domain name is exactly available; there is an expiration date, but even if you take into account the various grace periods, the domain will almost always not be deleted right at the exact hour/minute/second it says in the expiration date; to lessen the load a registry might for example take all domains supposed to be deleted (made available) on date X and spread them randomly during that day
- without saying too much, you can easily imagine that the registry infrastructure is like multiple inbound EPP servers to get all commands from registrars but then at some point there ought to be a single piece of code that forces the sequencing. The order of commands entering this "mono thread" will determine who wins. And there lies all the black magic. But it is easy to understand that before accepting an order a registry will have to check various things, besides the fact if the domain exists or not. If two orders for same domain come at the same time at the EPP level, but one order is more "complicated" than the other hence requesting more check work before going in the final stage, then this order may become second, and loose. Some registrars are smarter in those guesses, because otherwise registries will not divulge that.
- there are a lot of similar technical tricks. EPP (used by almost all registries) is based on XML. In XML, the node
<domain:name>can be semantically equal for all purposes to the node
<d:name>as long as the same XML namespace was tied to alias
d, so it makes no difference in the processing... but network wise, one case is 5 bytes smaller. Seems insignificant but multiplied across a single EPP frame and then all exchanges it can add up. Less to send means some advance and hence getting the answer faster. XML is open to many "improvements" like that, for example line returns are almost always not necessary. One less character... And so on.
Note also that all of this is moot in some TLDs that work with a pre-registration system, where registrars can send a command before expiration to ask to get the domain on deletion. There is no rush then. But the difficult problem to decide how to handle multiple requests for the same name. And the money part of it.
Also the above is just small part of the whole thing, as it is just about the technical link between registrars and registries, the part that end customers do not see. On top of this is the backordering system shown by registrars to customers. It can have various forms, and registrars are not bound by any rule to be equal among customers, so they can be auctions and any other systems.
To further elaborate on the selected answer, the famous dropcatchers typically own hundreds of registrars, to increase the number of connections to the registry (eg: Verisign) and increase the odds of catching names. Those registrars are actually shell companies.
See the list of registrars here: List of Accredited Registrars
Type Dropcatch in the combo box, for example and see how many sibling companies exist: DropCatch.com 1000 LLC, DropCatch.com 1001 LLC, DropCatch.com 1002 LLC, etc
Obvious maintaining an army of registrars is not cheap. There has been some consolidation in the industry. Any accredited registrar can play the game, but statistically the one who has the more connections is going to catch more of the coveted names.
The registrars normally have to abide by certain rules (which I don't know exactly), like limiting the number of requests per seconds, or staying within certain bandwidth limits, or they get a penalty. If you DDOS the registry with your scripts, obviously you'll get throttled or banned.