Request to relax the maximum cardinality of dct:format and dcat:mediaType

[pennyl67]

In LanguageDCAT-AP, we use dct:format and dcat:mediaType with maximum cardinality ..n , while DCAT-AP sets the maximum cardinality for both to ..1.
In order to appropriately describe datasets such as annotated corpora, consisting of the source files and their annotations (cf. DankMemes Dataset with source images in JPG format and their annotations in CSV), or datasets composed of data files in various formats (e.g. Audio Skills including video files in MP3 and WAV), we need to allow for multiple values of formats added on the same Distribution.
Note that we have considered using different Distributions for the different formats but this is not deemed appropriate as (a) the semantics of dcat:Distribution are meant for multiple representations/serializations of the dataset and not for the different filetypes included in the dataset, and (b) we prefer to use multiple Distributions for the representation of the same Dataset with different licences. (e.g., offered for free for research purposes vs. on a fee for commercial applications).
We have also considered using a different property, but we would rather use the two properties above for interoperability purposes when exchanging datasets across data spaces and other catalogues.

[bertvannuffelen]

@pennyl67 in DCAT(-AP) there is the possibility to have a distribution as a package. See https://semiceu.github.io/DCAT-AP/releases/3.0.1/#Distribution.packagingformat.

In that case a distribution is a collection of files that could be shared as one.

It might not yet be used so often, because it is unclear what the "format" is when multiple kinds of e.g. audio formats are together. In that case I would advice not to provide a value for format.

[pennyl67]

Thanks for the reply @bertvannuffelen . Indeed we are also using the packaging format but this is used for the format of the package itself, e.g. zip, gzip etc.
We still need, though, to record the values for the formats of the individual files. This is important because, for instance, in workflows for processing data by services, we use the "file format" and "language" properties to match together datasets and processing services (e.g. a translation service that takes as input text files of English in plain text or xml format).