Why are ospf non-backbone areas connected to backbone areas?

Why are ospf non-backbone areas connected to backbone areas?
My colleague left the job the other day and asked questions during the interview. Why does one of the ospf non-backbone areas have to be connected to the backbone areas. This problem is quite classic and involves ospf loops. In ospf, lsa is identified by LS type, link state id, and adverting router. Data exchange between non-backbone areas must be performed through the backbone areas. The routes between regions are encapsulated into three types of lsa and sent to another region to transmit routing information through the API. In the lsa test, the link status information is no longer pure routing information. Because the d-v algorithm cannot guarantee the elimination of the route self-ring, the self-ring is generated because the ABR router that generates the route will not be added to the generator information, no route information is generated by the original user. When ospf generates the lsa, the user's router ID is added to the lsa. If the information is transmitted in more than two regions, the original user information is lost. The solution is to encapsulate the routing information in the region into an lsa and send the information to a specific region. Then, the region forwards the information to other regions and regions, each lsa knows exactly the information of the creator. transmission between non-backbone regions is sent to a specific region, so no route self-ring is generated.