Website ABCs >
2) Domain Name System in Action for Dummies
Once upon a time there were tribes of natives who wanted to communicate across vast lands. They decided to create line of site communication from hilltop to hilltop using reflective materials and lights. Imagine such a system spanning from Colorado to the Atlantic -- and not just in one straight line, but in a mesh of networks like our current interstate system. And then imagine the same type of mesh network from Colorado west. Imagine the highest point in Colorado as being a single node through which all data crossing from east to west had to pass, and imagine that was the place in which a master map of the network was maintained at all times.
Well, that is about the best visual you could have for the Domain Name System, except for we forgot to include the part about caching (pronounced "cashing") prior requests and our example is limited to the US when that is not close to accurate, but this is a good start...
Putting it to use
When an internet user in South Carolina who likes South Carolina mustard based BBQ wants to research his tomato based Texas competition, he/she might type in www.stubbsBBQ.com to the address line of her browser. When she hits enter, a scouting message goes out across the line of site network towards Colorado asking for a definitive numerical address (an IP Address) for www.stubbsBBQ.com. At each node (hilltop) in the network the question is asked, "do you know the IP Address for "www.stubbsBBQ.com" ? The node monitor doesn't have a map of the network at all, but they do have a list of requests and responses that have passed through their node for the past 24 hours and if they've seen that request and response come thru they will respond to the request. If they have not, they will pass the request to the next upstream node (hilltop) from them until it is answered (or until it gets to Colorado where all answers are known) . The intermediate nodes keep this information for 24 hours from the last request time to lower upstream network traffic. This system of "caching of information for 24 hours" is why it can take up to 24 hours for web hosting changes to be realized by someone seeking a website. While new web hosting address setting changes will make it to Colorado instantly, all requests for your information may not make it to Colorado if down stream nodes have old info -- and it could be as many as 24 hours until the downstream cached answers related to your information are flushed from the child node memory.
Once a final IP address is known for a given request, either at an upstream node or Colorado itself, that information is returned to the users computer and then another request goes out from the users computer onto the internet in search of an IP Address (as opposed to the name you sent out first), and a similar but different routing process transpires to get the request to Stubb's website and to get the text from his website back to your web browser for viewing.
Well, that is about the best visual you could have for the Domain Name System, except for we forgot to include the part about caching (pronounced "cashing") prior requests and our example is limited to the US when that is not close to accurate, but this is a good start...
Putting it to use
When an internet user in South Carolina who likes South Carolina mustard based BBQ wants to research his tomato based Texas competition, he/she might type in www.stubbsBBQ.com to the address line of her browser. When she hits enter, a scouting message goes out across the line of site network towards Colorado asking for a definitive numerical address (an IP Address) for www.stubbsBBQ.com. At each node (hilltop) in the network the question is asked, "do you know the IP Address for "www.stubbsBBQ.com" ? The node monitor doesn't have a map of the network at all, but they do have a list of requests and responses that have passed through their node for the past 24 hours and if they've seen that request and response come thru they will respond to the request. If they have not, they will pass the request to the next upstream node (hilltop) from them until it is answered (or until it gets to Colorado where all answers are known) . The intermediate nodes keep this information for 24 hours from the last request time to lower upstream network traffic. This system of "caching of information for 24 hours" is why it can take up to 24 hours for web hosting changes to be realized by someone seeking a website. While new web hosting address setting changes will make it to Colorado instantly, all requests for your information may not make it to Colorado if down stream nodes have old info -- and it could be as many as 24 hours until the downstream cached answers related to your information are flushed from the child node memory.
Once a final IP address is known for a given request, either at an upstream node or Colorado itself, that information is returned to the users computer and then another request goes out from the users computer onto the internet in search of an IP Address (as opposed to the name you sent out first), and a similar but different routing process transpires to get the request to Stubb's website and to get the text from his website back to your web browser for viewing.
Summary
The DNS system is in fact a very simple system. It just doesn't work quite the way one might think it would. The method of caching requests at nodes as opposed to a more formal hierarchy of authority servers is in fact very simple and interesting, and the fact that a node on the network only knows the next node in each direction as opposed to having a map of the master network is also interesting. The 24 hour caching example here is the "industry norm". In theory, there are ways to try to encourage dns servers to shorten their caching time period, but those methods don't always work as those managing the dns servers have final say for their caching protocals.