You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I would rather not proliferate standards we intend to deprecate. If we know BIDS-prov is coming down the pipeline, it seems counterproductive to me to standardize on something else.
So, I take it as if we are to converge on BIDS-prov, we should not have totally unrelated PROV records of different kinds in a BIDS dataset, and they should if not "absorbed" (like with GeneratedBy in #148), then "integrated" with, hence this issue for another type of "provenance" capture we already have in BIDS.
Is your feature request related to a problem? Please describe.
BIDS already has "descriptions" to provide freeform textual annotation and extra metadata for _desc- entities, and potentially some other columns with relevant options.
That is "freeform provenance" which might also correspond to some provenance contained in specific _prov.jsonld files at some level.
Describe the solution you'd like
I wonder, if somewhat similar to HED columns in _events. files we could gain some PROV column in *_descriptions.tsv file(s) linking those free form activities with activities formalized in "BIDS-Prov".
ATM, current proposal, if I got it right, proposes to duplicate (although with different activity UUIDs and date times) prov records from the same pipeline ran with the same options. That actually dilutes information and although would provide the ultimate PROV record for tools (and some humans) - would make it actually hard(er) to "grasp" what was done for a specific file. May be even if not "piggy backing" on inheritance via json files (#146) and having PROV record within JSON, we could have some _prov.jsonld files which would provide reusable specifications relating to _descriptions.tsv somehow?
PROV column in _descriptions.tsv could contain the id (or some other marker to be found in .jsonld files) for an activity detailed with the parameters etc shared at some level in a _descriptions_prov.jsonld file, potentially without pointing to specific filenames for input/output etc. So, such file by itself would not be a complete provenance record, but rather "distill" of processing pipeline common to all files with that _desc- .
Describe alternatives you've considered
Additional context
The text was updated successfully, but these errors were encountered:
This is a bit raw... But my thinking is inspired by @effigies comment elsewhere:
So, I take it as if we are to converge on BIDS-prov, we should not have totally unrelated PROV records of different kinds in a BIDS dataset, and they should if not "absorbed" (like with GeneratedBy in #148), then "integrated" with, hence this issue for another type of "provenance" capture we already have in BIDS.
Is your feature request related to a problem? Please describe.
BIDS already has "descriptions" to provide freeform textual annotation and extra metadata for
_desc-
entities, and potentially some other columns with relevant options.That is "freeform provenance" which might also correspond to some provenance contained in specific
_prov.jsonld
files at some level.Describe the solution you'd like
I wonder, if somewhat similar to
HED
columns in_events.
files we could gain somePROV
column in*_descriptions.tsv
file(s) linking those free form activities with activities formalized in "BIDS-Prov".ATM, current proposal, if I got it right, proposes to duplicate (although with different activity UUIDs and date times) prov records from the same pipeline ran with the same options. That actually dilutes information and although would provide the ultimate PROV record for tools (and some humans) - would make it actually hard(er) to "grasp" what was done for a specific file. May be even if not "piggy backing" on inheritance via json files (#146) and having PROV record within JSON, we could have some
_prov.jsonld
files which would provide reusable specifications relating to_descriptions.tsv
somehow?PROV column in
_descriptions.tsv
could contain theid
(or some other marker to be found in .jsonld files) for an activity detailed with the parameters etc shared at some level in a _descriptions_prov.jsonld file, potentially without pointing to specific filenames for input/output etc. So, such file by itself would not be a complete provenance record, but rather "distill" of processing pipeline common to all files with that_desc-
.Describe alternatives you've considered
Additional context
The text was updated successfully, but these errors were encountered: