Closed msporny closed 12 months ago
Do we make it invalid for someone to do this: the
@base
JSON-LD directive is ignored in an external context file, but it is valid this way...
Hrm, what we don't want to do is allow the BASE to be set by the processor based on where the document lives, IIUC. So, setting @base
in the way that you showed in your example is fine -- because the BASE won't be mis-interpreted based on location, which is what this PR is attempting to fix. That said, processors that run in non-JSON-LD mode won't process that @base
definition, leading to potential interop issues, so perhaps the guidance in this PR is best in order to increase interop.
Do we make it invalid for someone to do this: the
@base
JSON-LD directive is ignored in an external context file, but it is valid this way...Hrm, what we don't want to do is allow the BASE to be set by the processor based on where the document lives, IIUC. So, setting
@base
in the way that you showed in your example is fine -- because the BASE won't be mis-interpreted based on location, which is what this PR is attempting to fix.
Ok, I understand
That said, processors that run in non-JSON-LD mode won't process that
@base
definition, leading to potential interop issues, so perhaps the guidance in this PR is best in order to increase interop.
Well, this PR does not cover this specific case. A non-JSON-LD processor will have no idea what to do with @base
, and it might simply ignore it. But then there is a discrepancy. My proposal would be that, beyond the text you propose in the PR, we should disallow the @base
statement for VC-s (I do not think it is frequently used, so it is not a major loss).
Well, this PR does not cover this specific case. A non-JSON-LD processor will have no idea what to do with @base, and it might simply ignore it.
Yes, exactly, which is why we should probably tell authors that they "SHOULD NOT use @base
".
But then there is a discrepancy. My proposal would be that, beyond the text you propose in the PR, we should disallow the @base statement for VC-s (I do not think it is frequently used, so it is not a major loss).
I have raised https://github.com/w3c/vc-data-model/issues/1266 to track this issue.
Normative, multiple reviews, changes requested and made, issues raised for remaining concern, no objections, merging.
This PR attempts to address issue #180 by ensuring that the base URL is set to null when processing as RDF.
Preview | Diff