Further clarification on (partial) resolution

[andrewwhitehead]

In the Read (resolve) section there is this paragraph:

On completing the processing and successful verification of all entries in the DID Log, respond to the DID resolution request, including the application of DID query parameters such as ?versionId= and ?versionTime= with the appropriate DIDDoc version and content.

While I don't disagree with this there are a couple side effects to consider.

• Is it the rule that a more recent, invalid log entry will prevent resolution of an earlier versionId/versionTime? ie. Users would no longer be able to resolve a previous valid entry because the log itself appears corrupt. This seems like a good policy from a security perspective.
• I think it might be also relevant when defining the behaviour of a cache, because the entire log file would always need to be processed/verified instead of needing to account for partial verifications. We would also need to define the expiry of a corrupted log, so that an unauthenticated/unwitnessed attack on a log file could be recovered from.
• The witness verification algorithm would also depend on this decision. It becomes more complicated if parsing of the log file might abort early.

[brianorwhatever]

Is it the rule that a more recent, invalid log entry will prevent resolution of an earlier versionId/versionTime? ie. Users would no longer be able to resolve a previous valid entry because the log itself appears corrupt. This seems like a good policy from a security perspective.

I don't think this should be the behaviour as there is nothing invalid about that previous entry

I think it might be also relevant when defining the behaviour of a cache, because the entire log file would always need to be processed/verified instead of needing to account for partial verifications. We would also need to define the expiry of a corrupted log, so that an unauthenticated/unwitnessed attack on a log file could be recovered from.

yeah definitely.. however less of an issue if we allow early resolution (previous entries aren't invalidated)

The witness verification algorithm would also depend on this decision. It becomes more complicated if parsing of the log file might abort early.

Witnesses shouldn't be asked to sign an earlier version in fact we should disallow it. so for them processing should never abort early