What's also coincidental is that I read a Quora post recently of network model databases where references are direct. They have the linking problem which CODASYL worked towards solving. Where you need to update all links when there are changes. My idea 18 data synchronization system is to produce one system to handle it. It also could be used for cache invalidation which is an interesting problem in itself.
Your solution also reminds me of a graph database where nodes and vertices/ edges are explicitly stored but in your case you are using reference counting which is new to me.
Thanks for sharing your knowledge on this.
From a logical point of view, explicit storage of links should be more efficient than a hash join or nested loop join. But is there a tradeof in write amplification? Joins in Postgres are materialised at query time but they're efficient due to btrees, whereas in your model the data links are all materialised. I think links in graph databases such as neo4j and dgraph are materialised as links directly too.
What's also coincidental is that I read a Quora post recently of network model databases where references are direct. They have the linking problem which CODASYL worked towards solving. Where you need to update all links when there are changes. My idea 18 data synchronization system is to produce one system to handle it. It also could be used for cache invalidation which is an interesting problem in itself.
Your solution also reminds me of a graph database where nodes and vertices/ edges are explicitly stored but in your case you are using reference counting which is new to me.
Thanks for sharing your knowledge on this.
From a logical point of view, explicit storage of links should be more efficient than a hash join or nested loop join. But is there a tradeof in write amplification? Joins in Postgres are materialised at query time but they're efficient due to btrees, whereas in your model the data links are all materialised. I think links in graph databases such as neo4j and dgraph are materialised as links directly too.