[FIX] URI "definition" and recommendation

[Remi-Gau]

TO DO

• add URI definition after the "File Formation specification" section
• make sure DOIs in the specs (in json examples or otherwise) have the format doi:<path>

For reviewers

• is this the right "place" for that definition?
• anything else but the DOIs that should change?

[sappelhoff]

anything else but the DOIs that should change?

IMHO all JSON fields that contain URLs or DOIs. For example CogAtlasID, which says:

source: table in https://bids-specification.readthedocs.io/en/latest/04-modality-specific-files/03-electroencephalography.html#sidecar-json-_eegjson

the big benefit would be that we can validate for URIs, see: bids-standard/bids-validator#1072

the drawback would be that probably so far people have been specifying DOIs either as a URL, or potentially using just the DOI without any prefixes.

But I could live with that drawback in order to make the spec better in the future. Could be considered a breaking change though 🤷‍♂️

[Remi-Gau]

This looks ready to go. Some changes are I think not problematic but some other I am much less sure about especially in the "higher level" files ("common principles" and "modality agnostic") as they could "break" things.

Will add comments through the review panel.

Once again will leave the linting as a last step when we agree content.

[Remi-Gau]

One last question: that relates to the diversity of format we use to refer to papers in the specs themselves.

There are a couple of links to some papers in the specs that use a URL rather than a DOI. Should those be changed?

See here:
https://bids-specification.readthedocs.io/en/stable/99-appendices/08-coordinate-systems.html#introduction

For example, should this:

(see e.g.,
[https://doi.org/10.1016/j.neuroimage.2012.01.024](https://doi.org/10.1016/j.neuroimage.2012.01.024))

be turned into this:

(see e.g.,
[doi:j.neuroimage.2012.01.024](https://doi.org/10.1016/j.neuroimage.2012.01.024))

[sappelhoff]

There are a couple of links to some papers in the specs that use a URL rather than a DOI. Should those be changed?

I think that'd be reasonable. And I think @yarikoptic would agree as well. See his comment: #403 (comment)

[Remi-Gau]

This one has been rebased and conflicts have been resolved.

So this should be ready to merge before we can move on to #628 for more of the same (and potentially #636 ?)

[sappelhoff]

LGTM!

However, we need other people to chime in about the potential breakage of now wanting our DOIs specified as URI in the from doi:thedoi

From a technical validation perspectives, only the datasets will break that specified the DOI as just thedoi, because those that used https://doi.org/thedoi still have a valid URI.

here's the link to previous discussion on this: #629 (comment)

[effigies]

This looks great! I have made some suggestions to maintain backwards compatibility. Let me know what you think.

[effigies]

I'd still like to have a general note with the URI definition, but I can be argued out of it if you're so inclined.

[sappelhoff]

Thanks @Remi-Gau

[yarikoptic]

There are a couple of links to some papers in the specs that use a URL rather than a DOI. Should those be changed?

I think that'd be reasonable. And I think @yarikoptic would agree as well. See his comment: #403 (comment)

yes, I do. Thank you! ;)