Replicated and Authenticated Information Servers

next up previous
Next: References Up: Areas Needing Development Previous: More Flexible Client

Replicated and Authenticated Information Servers

We have observed the following problems with network information servers:

  1. Internet servers are frequently overloaded or unavailable for various reasons.
  2. Servers currently provide no assurances of authenticity or integrity.

To deal with the above problems, a universal location-independent naming scheme needs to be coupled with file replication and authentication mechanisms. Transparent replication via a name-to-location lookup service will achieve fault tolerance and load balancing for end-user network services. File servers should be required to authenticate themselves before being allowed to replicate files and update the name-to-location mapping. A description associated with each location-independent filename should provide an MD5 or similar fingerprint for verifying the integrity of a retrieved file.

Jack Dongarra
Sun Dec 18 14:22:28 EST 1994