pdsinterop / solid-link-metadata Goto Github PK

As noted on Gitter by @gibsonf1 (thank you!)

[..] There is a scale issue as well. For example, we have a service that converts a pdf into units of thought which results in up to milions of triples per pdf. So it could be a big burden for a server to handle redirects etc for millions and millions of uris per moved user

A resource in our case is as small as a single triple, and there can be 100's of millions of these in a pod, so scale is an issue but I guess you really just need a pattern matching at the server level to do it. Such that any request with a certain uri pattern would redirect to another pattern.

This should either be discussed and addressed in the spec, or mention may need to be made if the spec is not otherwise adapted in this direction.

As part of the feedback on the spec and vocab on the Solid forum, a scenario was mentioned that we currently do not address explicitly (although it is implied across the various handled use-cases).

We may want to keep a copy at the source, instead of only having a copy at the target

As redirects may be temporary, not permanent, it is not required to delete the original file. Adding a meta entry at the source is sufficient.

This needs to be added in more detail to the existing use-cases to clearly describe the distinction with the implicit mentions.

There is an issue in the Solid Spec repo that discusses redirecting and remapping.

For those interested, visit: solid/specification#136

Various points from that discussion can/should be added here, either directly in "our" spec or as an issue in this repo.

We should take care to make sure any changes here are also reported there (as well as on Gitter chat but that usually sorts itself out).

There is also a direct request/question whether "our" spec could be adopted by the main spec: solid/specification#474

In several places in the spec the word you is used instead of the relevant actor (e.g., server, client, etc).

you should be changed for all occurrences where it is not addressing the reader.

Reported on Gitter by @csarven (thank you!)