We ask for filename in the Content-Disposition but we don't do anything with it

[richard-jones]

During deposit, filename is requested in the Content-Disposition header, when sending a binary/package. We then don't place any onus on the server to do anything with this information, and we don't provide any mechanism for this information to be included in the status document.

Is filename even meaningful? It MAY affect the @id that's given to the file, but it may also not.

What should we do? Get rid of filename in Content-Disposition, keep it and just document it's usage, or make space for it in the status.link field?

[alexdutton]

FWIW, invenio-sword uses the filename to set the file key (i.e. the filename), which is then used in constructing the @id.

AIUI, the spec effectively says a deposit is a set of identifiable but unstructured files, with neither a directory structure, or even filenames. This probably wants resolving, even if every implementation is going to pay heed to filenames anyway.

[richard-jones]

This is kind of a duplicate of #9 - I've left it that the spec allows the client to specify it, and that the server SHOULD take it into account, with no requirement to do so. On update the server MAY do something with it but may not. So this is just a hint to the server that this is what you would like to call the file. I don't think I'm going to say any more about it in the spec, though @alexsdutton this may be something for the implementers guides?