split centres.xml across the individual recommendations?

[bansp]

I am not sure if we can find serious arguments against keeping the centre info separate from the recommendations.

Currently, we list 51 centres and have 52 recommendation documents, and that's already a yellow blinking light.

Each recommendation tells us something about the single centre, they are connected implicitly, via the file name. So, the info on the IDS in centres.xml is:

  <centre id="IDS" deposition="1">
    <name>Leibniz-Institut für Deutsche Sprache</name>
    <a href="https://centres.clarin.eu/centre/11"/>
    <nodeInfo>
      <ri status="B-centre">CLARIN</ri>
      <ri status="Collections, Lexical Resources, Operations">Text+</ri>
    </nodeInfo>
  </centre>

... and the filename of the recommendations is IDS-recommendation.xml. The recommendation file has a header, currently set to the following (and note the redundance between the file name and the filter element):

  <header>
      <lastUpdateCommitID>76e7d21871f644d0ea8b5b62811a493304d9cd18</lastUpdateCommitID>
      <filter>
         <centre>IDS</centre>
      </filter>
      
      <respStmt>
         <curator>Piotr Banski</curator>
         <github>https://github.com/bansp</github>
         <reviewDate>2023-07-24</reviewDate>
      </respStmt>
      
   </header>

(where the maintainer info was added for the sake of a demo at CAC-2023 and I don't consider it binding...)

Arguments for merging:

• data modelling: these pieces belong together, the relationship between the centre element and the relevant recommendation file should always be 1:1
• ease of maintenance: we'd only expect PRs of the recommendation files
• logic / data modelling: centres info is secondary to recommendations; it is not independent; we are not a centre-information facility -- that is the role of the CLARIN DB and the databases of other research infrastructures; we should only query those DBs in a sanity-checking operation, once in a while
• again a bit of a data-modelling issue: IF, say, centre X recommends one set of formats as a CLARIN/Text+ centre, and another as a DARIAH centre (and it's really a single institution, and that's how we want to represent it) then we can allow RI-specific attribute on <formats>: <formats ri="dariah"> would be DARIAH-specific while <formats> would go for all the RIs involved. (this is not yet technically possible and won't be until we see the need for such an arrangement; but it would be good to have the base for that in place)
• centres list is a projection of recommendations; we don't use it separately for the (currently somewhat dormant) standards information part; it doesn't make sense to say that one centre does or does not use the given standard; the standards information should exist separately, and be information on standards, with maaaaybe a mention of some centre in the overall description. We don't need a skeletal list of centres, used by both format recommendations and standards info.

Argument against: technical complexity and bugs cropping up after the rearrangement, but that's a non-argument, really. It might be an argument for releasing 3.0.

[bansp]

To be sure: I suggest that each <centre> element goes into the header of a recommendations file, and that we get rid of centres.xml.

Being able to use the attribute @ri on <formats> is one thing that would follow. Another would be the use of that attribute on <name>, to differentiate between the centre name used for one RI vs. another (this is with reference to issue #245 ).

[bansp]

Ah, incidentally, the <a> element is extremely anti-semantic, we should get it replaced, probably with added flexibility for RIs.

[margaretha]

Eliza will move centre elements to the recommendation files.

[bansp]

I will add the element as optional, for the time being, to the schema for recommendations. After the move, I'll make it obligatory.

[margaretha]

I mean in in export files when the format recommendations are not filtered by centreID, e.g.

<format id="fComa">
   <centre>HZSK</centre>
   <domain>Contextual Information</domain>
   <level>recommended</level>
</format>

[margaretha]

I have added centre elements in the recommendation files and updated the code to use them instead of the centres.xml. Please check the dev branch.

[bansp]

Ah, I was wondering when I saw the duplicated info, both in centres.xml and in the individual recommendations. And it's been waiting for me! :-)
Well, let it wait a while longer, since we have the issue open still so it won't be forgotten.
To be sure: the move looks fine to me, but I've only had a look rather than giving a try to a version with centres.xml either thinned down or altogether removed.
I'll get to that in mid-May, I expect.

[bansp]

It's mid-May so I am going to have a look now and most probably close it.

[bansp]

Don't want to close the ticket yet, because the info is still duplicated. I suggest that we remove centres.xml after the upcoming release and then deal with potential consequences. I'm moving this ticket to 2.8.0; it already has the priority label.

Edit: oh let me wait with the milestone switch until tomorrow (May 24th). Maybe we can remove centres.xml right away.