What if my content includes files called inventory.json or inventory.json.sha512?

[zimeon]

I see two options:

1. With the spec as currently written there would need to be a note explaining that these are special cases where the existing file path cannot be the original filepath, it must be renamed inside the OCFL object, e.g.:

[object root]
|- v1
|    |- inventory.json   <-- the actual inventory, not the file so named in v1 content
|    |- inventory.json.sha512   <-- the actual digest sidecar
|    |- inventory.json_moved   <-- the content called inventory.json in v1 state
|    \- inventory.json.sha512_moved  <-- the content called inventory.json.sha512 in v1 state
\...

2. Use another directory as is done in BagIt (where it is called data) so that there is clean separation:

[object root]
|- v1
|    |- inventory.json   <-- the actual inventory, not the file so named in v1 content
|    |- inventory.json.sha512   <-- the actual digest sidecar
|    \- data
|          |- inventory.json  <-- the content called inventory.json in v1 state
|          \- inventory.json.sha512  <-- the content called inventory.json.sha512 in v1 state
\...

[ahankinson]

I think this is a problem, but I'm not sure how big a problem it is.

1. The inventory files are only valid in the root of the version directory. Any other inventory files are treated as content.
2. Inventory files in the version directories are optional (but recommended)

If we have to address it, though, I would go with option 2.

[ahankinson]

Maybe call it content, instead of data, since that's what we seem to be calling it anyway?

[zimeon]

I agree that if we were to adopt option 2 then content would be a better name.

[neilsjefferies]

...or make the name ocfl_inventory.json etc. to make it a minimal chance of collision and just say they are reserved and you can't have them.

[ahankinson]

@neilsjefferies that might get a bit messy if we're wanting to handle arbitrary content from someone's HD. Especially if they've adopted OCFL as a way of organizing their content. :)

[neilsjefferies]

That would be archivists just being silly. If that is what you are doing then packaging is really the way to go.

[awoods]

It is probably important that we support the case of someone's content having the same name as OCFL administrative metadata files.
That being the case, I would also prefer option no.2 for its avoidance of renaming files.

[zimeon]

I lean toward having an extra content directory to make it clean/clear

[neilsjefferies]

Since the inventory already has a mechanism for separating logical path in the object from actual path on disk, isn't this really just an Implementation Note. Since you can't have those names on disk, rename them - they can still have the name in the inventory. The system we have can handle it already, why introduce a change for a small corner case?

[ahankinson]

TBH, having the inventory in the same directory root as content has never really sat well with me. I just wasn't able to say why. It seemed like we were mixing administrative and content data.

[ahankinson]

@rosy1280 @julianmorley pretty please could we have some input on this so we can move it along?

[rosy1280]

i think it would be cleaner and more human readable if their was a content directory and we put the content files in them. there is a reason that bagit and moab do this. so lets not reject use cases because we don't think people should do them.

[julianmorley]

Yeah, having a content directory makes a great deal of sense. It enables a clean nesting of OCFL objects inside OCFL objects.

[rosy1280]

also when you start to think about distributed digital preservation (which i don't want to broach, but...) if Emory's repository implements OCFL and sends their content to Chronopolis that has implemented OCFL you may end up with clashing inventory.json and inventory.json.sha512

[zimeon]

Per https://github.com/OCFL/spec/wiki/2018.10.17-Editors-Meeting agreed to use content directory

[ahankinson]

if it wasn't clear by the assignments, I'm currently working on a PR to this effect

[ahankinson]

Q: Do the 'logical' filepaths also omit the 'content' part?

[awoods]

@ahankinson : it would make sense to me that content would be excluded from logical file paths.

[ahankinson]

👍

[ahankinson]

Also, are content directories a MUST?