Open kir0ul opened 1 year ago
bendichter
commented
1 year ago @kir0ul it sounds like you might have a lot of repeats of the same issues. If they are critical, it's probably a good idea to at least take a look a them. Would you mind copying/pasting the output of NWB Inspector?
kir0ul
commented
1 year ago Hi @bendichter, so we were able to fix all the critical issues and push the updated dataset to DANDI. But I was more hoping to raise the general issue of updating a dandiset which has already been validated and accepted. I guess this updating issue may hopefully become less of an issue with time, as NWBInspector and its DANDI config converge and get less changes in the future.
bendichter
commented
1 year ago Yeah, it's a fair point. We have back doors for dealing with this if we need to but I'd rather at least try going through the front door first 🙂
yarikoptic
commented
1 year ago Yes, I understand the concerns about the increasing strictness of standards and validators over time. While it's beneficial in ensuring higher quality and better organization of data, it can indeed pose challenges for those who need to adjust their existing data to new norms. It's important to remember that the ultimate aim is to improve data and metadata quality for the sake of better discoverability, usability, and overall value.
In this context, it could be insightful to identify and discuss the 256 issues that have been encountered, especially the 106 that are classified as critical. By understanding what these issues are, we can find ways to tackle them effectively and efficiently, minimizing the burden and stress associated with compliance.
Looking ahead, we could consider engaging previous data submitters to update their datasets if we discover critical defects in their data or metadata. This proactive approach is intended to enhance the data quality, making it more beneficial for all users. We need to communicate this purpose clearly to avoid creating unnecessary pressure or misunderstanding among contributors. It's not about pointing out faults, but fostering a collaborative effort to improve data quality and usefulness.
kir0ul
commented
1 year ago Here are the logs:
Proposed change
Our use case is that we published a dataset on DANDI several years ago $\rightarrow$ now we need to update this dandiset a few years later, but the validating rules have changed a lot since then. The proposed change would be to allow updating an already published dandiset with the original version of NWBInspector which has been used at the time of publication.
Alternative options
Fix all the issues listed by NWBInspector, but it's quite a burden to make just a small update. Here we are talking about 256 issues (106 critical ones). If the cost to update a dandiset is too high, the risk is that people may decide to leave it as is without correcting the information that need to be corrected.
Who would use this feature?
Anyone who may have to update a dandiset after some time.
(Optional): Suggest a solution
One solution could be to save in the dandiset the version of NWBInspector which has been used when the dandiset has been published originally, and validate the updated dataset against this specific version of NWBInspector.