NOTE: the following is my generalized overview of some thoughts I came up with months ago in regards to LSID resolvers, and how to architect a fault tolerant solution (LINK). I missed the meeting in Denmark last week (I was on a family vacation for once) where they were discussing this, and wrote the following for another attendee to submit on my behalf. I’m posting it here for further exposure and discussion of the merits and shortcomings of these ideas.
“In thinking about the architecture for future LSID resolvers we need to remember that a single point of failure will fail. This has been proven true too many times, and it’s clearly not the way to proceed if we want to build a system the community can rely on. To succeed we should follow successful implementations of distributed software services such as the domain name server (DNS), and the network time protocol (NTP). These succeed because the protocols are standardized and the software to connect to, and utilize, these services are simple to deploy. If we had a package-able server that would connect to these services and be easy to deploy, an institution could ultimately count on its own server (with other servers providing automatic fail over) to serve as a resolver that would be updated against lead node servers regularly.
Home