My preference, now that subreferencing exists, would be to converge on a future where we have some FRBR-style clustering of sources (similar to what fatcat.wiki does for the range of different URLs / content hashes that refer to the same source, but compounded with the range of different ways it can be cited), which assigns CIDs, and lets curators explicitly merge/split CIDs where needed. Then we end up with one CID for each cluster of substantively identical sources.
This matches what any meta-analysis of citation usage, citation affect, source reliability, retraction, &c would want to make bidirectional source analyses useful. It's a bit of extra structural work up front, and would result in a CID namespace of O(100M) cited sources, but would be a clear and new public good useful immediately to our editors and to reuse and analysis beyond our projects. I'm not sure what that means in the *immediate* future, but just as fatcat did when choosing its battles, a recognition that we will start using a CID namespace that supports future merges would let us start small and potentially refactor initial uses via merges into any slightly different future implementations. On Thu, Aug 22, 2024 at 3:41 PM Strainu <strain...@gmail.com> wrote: > I hit the same problem recently. My solution was very similar to what Adam > proposed: build the same ref again (completely) and calculate the hash. > Since in rowiki all interaction with Wikidata is through modules, it was > trivial to extract the reference code and invoke it directly or from a > template. > > One nasty problem that I couldn't really solve nicely was that the CS1 > module would add a templatestyle which on subst would be expanded to a > different strip marker for each instance, causing "reference with same name > and different content" errors. This meant I could use a template, but not > substitute it. If I understand your use case correctly, you won't have this > problem. > > When this feature becomes available, you could simply adapt the template > to generate the <ref extends> tag. > > Hth, > Strainu > > Pe miercuri, 21 august 2024, Adam Wight <adam.wi...@wikimedia.de> a scris: > > Hi, I'm one of the developers working on this project. > > Thanks for the question! Currently, the state of our thinking around > reusing refs from a template is that it's problematic, but the simplest and > safest existing workaround is to produce another ref in exactly the same > way as the first. So if the CS1 family of templates is used (eg. Cite book > / Literatur) then you can use the same or a similar template to generate > the second ref, and if the parameters to CS1 match then your final refs > will be identical and the footnote markers will be merged together into one > symbol and one reflist item. > > The second possibility would require a feature change which has some > questionable implications: that any two refs with identical content could > be merged in reading mode, regardless of the "name" attribute. However, > since the reference content can only be guaranteed to be identical if the > same template is used, I believe the first solution (use the same > underlying template and let it produce a name) is the best workaround for > the moment. > > Please do suggest any better ideas you might have... so far, it's been > hard to imagine what a more "stable link" might look like, maybe it's an > explicit "cid" or HTML id, we're really not sure yet. > > Regards, > > [[mw:User:Adamw]] > > > > On Tue, Aug 20, 2024 at 10:47 AM Bináris <wikipo...@gmail.com> wrote: > >> > >> I have just experienced a related problem recently. I am not sure if it > is in the scope of your project, so I just mention it here. > >> Original problem: a reference is given in Wikidata as source of death, > for example. It appears automatically in infobox. > >> Later in the article I need that for another purpose, and it will > appear on tha pege twice. > >> Question: can we make it appear once? > >> Answer in huwiki: the link of Wikidata contains a hash, which can be > used as <ref name="this-hash"/>. > >> New problem: whenever the source is edited in Wikidata, the hash > regenerates, and the article will silently be spoiled. We can discover the > error in ref name only accidentlly, and it is a head ache to guess the > original. > >> Relation to this project: can we safely reuse the references > originating from Wikidata? Can it offer a stable link? > >> Or should I write this to Wikidata mailing list? > >> _______________________________________________ > >> Wikitech-l mailing list -- wikitech-l@lists.wikimedia.org > >> To unsubscribe send an email to wikitech-l-le...@lists.wikimedia.org > >> > https://lists.wikimedia.org/postorius/lists/wikitech-l.lists.wikimedia.org/ > > > > -- > > Adam Wight - Developer - Wikimedia Deutschland e.V. - > https://wikimedia.de > > _______________________________________________ > Wikitech-l mailing list -- wikitech-l@lists.wikimedia.org > To unsubscribe send an email to wikitech-l-le...@lists.wikimedia.org > https://lists.wikimedia.org/postorius/lists/wikitech-l.lists.wikimedia.org/ -- Samuel Klein @metasj w:user:sj +1 617 529 4266
_______________________________________________ Wikitech-l mailing list -- wikitech-l@lists.wikimedia.org To unsubscribe send an email to wikitech-l-le...@lists.wikimedia.org https://lists.wikimedia.org/postorius/lists/wikitech-l.lists.wikimedia.org/
