Storage versioning

[rullzer]

A more detailed plan after #9445 and now that #12129 is in we can look at this in a bit more depth.

I think it is important to make a distinction here. The implementation in #12129 is to allow version backends. Which is why we have versions in group folders for example.

This is about adding versioning support on a storage level.

I'd propose a new interface to be implemented by storages. IVersionStorage. This should at least have a method versioningEnabled (or something similar). Because it might not always be available (e.g. S3 versiniong).

Then this would need the following methods I think

• getVersions(string $path): IVersion[]: obtains a list of the current versions
• createVersion(string $path): creates a new version (this can also be a void method if that happens automatically on write?)
• restore(IVersion $version): bool: restores the version

Then IVersion needs

• size
• time
• id
• mime?

That way we could use it in combination with #12129

@icewind1991 for feedback.

[icewind1991]

There is no need for a new storage interface, with the new modular versions (and trashbin api), an app can register a version backend for a specific storage class.

#12129 should already have all the interfaces and core parts needed to implement storage versioning

[rullzer]

@icewind1991 well it kind of depends. Some storages (the obejct storage for example) are not apps.

Also we'd then need to think how to handle layered storages. As an example. Take groupfolders. In the end you'd also want those versions to end up on the versioning of the storage.

[icewind1991]

if it's not an app the backend can just be registered in the versions app with the default backend.

For the layered storage we can try and get the versions manager to pick the s3 backend over the groupfolders backend (the s3 backend should behave correctly for the groupfolders usecase anyway)

[MorrisJobke]

For the layered storage we can try and get the versions manager to pick the s3 backend over the groupfolders backend (the s3 backend should behave correctly for the groupfolders usecase anyway)

@icewind1991 Is this done?

Can we close this ticket here?