I think a central problem here is that the "dimension of rawness", unlike a memento timestamp, is really a property of the web archive, and not of the memento.
It seems the correct way to present dimensions of rawness seems to be when querying information about a web archive. The https://github.com/webrecorder/public-web-archives project presents a way to define different replay options available, per web archive.
A web archive either contains and is able to serve raw content, or it doesn't. If it does, and if it modifies content, then it can serve raw and rewritten content for each memento. It is hard to foresee, especially given the example dimensions of rawness, why an archive would for instance have original-content for some mementos, and not for others.
Even if there was such a use case, a client must still have a priori knowledge of the possible dimensions of rawness.
For example, if a client tries Prefer: foo, and the response does not include this preference, does that mean:
The archive doesn't support foo just for this Memento, but perhaps for others.
The archive doesn't support foo for any Memento
foo is not a valid preference to be applied
A concise manifest of possible preferences/dimensions of rawness per archive can addresses all of these issues.
I think a central problem here is that the "dimension of rawness", unlike a memento timestamp, is really a property of the web archive, and not of the memento.
It seems the correct way to present dimensions of rawness seems to be when querying information about a web archive. The https://github.com/webrecorder/public-web-archives project presents a way to define different replay options available, per web archive.
A web archive either contains and is able to serve raw content, or it doesn't. If it does, and if it modifies content, then it can serve raw and rewritten content for each memento. It is hard to foresee, especially given the example dimensions of rawness, why an archive would for instance have
original-contentfor some mementos, and not for others.Even if there was such a use case, a client must still have a priori knowledge of the possible dimensions of rawness.
For example, if a client tries
Prefer: foo, and the response does not include this preference, does that mean:foojust for this Memento, but perhaps for others.foofor any Mementofoois not a valid preference to be appliedA concise manifest of possible preferences/dimensions of rawness per archive can addresses all of these issues.