HCLSIG BioRDF Subgroup/Tasks/URI Best Practices/Recommendations/AttitudeTowardMigration

From W3C Wiki
Jump to: navigation, search

[[/../../Recommendations|URI Note main page]]

Attitude toward migration

The overall problem is that people mint URIs that point to their own web sites, and these URIs work for a while, but then the project is cancelled, the author changes jobs, etc., and the heir to the URIs fails to maintain the resources.

The meta-problem is what to say about this in the URI note. We propose simple conventions for doing client-side redirections (see [[/../URI_Resolution]]), and will make efforts to get agents to use them. But do we say that ephemeral URIs are OK (not good, but OK), because the resolution rules will deal with them, or do we tell people that to be a good citizen they need to go out of their way to get ahold of persistent URIs, such as PURLs?

Status quo

Network resource migration is not implemented in the HTTP scheme except via changes DNS records. There is no way to migrate only some of a web site's resources without using HTTP redirection, which involves the active participation of the original server.

Migration is currently handled manually. A 404 is observed by a site administrator, and the broken link gets removed or replaced by a link to the same or similar material residing at a different location.

The LSID protocol has several mechanisms for handling migration.

In defense of ephemeral URIs

Some resources are going to move. This is inevitable. If you think they mustn't you're being unrealistic.

RDF is important enough to put in archival (read-only) locations.

Truly durable URIs are very difficult to achieve. They will inevitably have a very high barrier to entry. Many useful projects don't use anything like purls, and reside in ephemeral locations. It is unreasonable to expect all references to them to be updated when they move.

Client-side redirection is useful for many reasons, including performance, privacy (access to enterprise-internal copy of something can't be observed outside), and connectivity (disconnected access). So if we need this anyhow, why not use it to redirect for the purposes of moved content?


This is the semantic web. We use web standards and web protocols and all that, not some ad hoc resolution thing that does an end run around the web. Using a different resolution method will break all sorts of things. (? anyone believe this ?)