Category talk:CS1 errors: access-date without URL

This category is being incorrectly added to citations which generate URLs through templates

In Australia, we have the problem of government agencies changing domain names or other parts of URLs from time to time, e.g. census data, gazetteers etc. So we use templates to avoid the use of static URLs which would be a massive task to replace manually. An example of such an article is Algester, Queensland which appears incorrectly in this category. When you look at the references in the rendered article you will see citations [1] and [2] do have URLs and do have accessdates but in both cases the URL is generated via a template. I think this category should not include articles which contain citation URLs generated by templates. Kerry (talk) 22:12, 13 October 2015 (UTC)

The access-date error and categorization in this category is caused by reference 3 because it does not have a |url= parameter but does have an |access-date= parameter.

—Trappist the monk (talk) 23:26, 13 October 2015 (UTC)

DOIs instead of URLs

This category also contains articles with citations where ther is a DOI but no URL, implying that this is an error that needs to be fixed. But unless there's something I'm missing, access dates are relevant for any online resource identifier, be it URL or DOI, so pages like that shouldn't really be categorised here. Any thoughts? Uanfala (talk) 18:55, 9 June 2016 (UTC)

You are not the first, and likely not the last, to wonder about this.

|access-date= is and always has been associated solely with |url=. Were it the way you suggest, and were there multiple identifiers (|doi=, |pmc=, |pmid=, |zbl=, ..., all of which make external links), to which of them would you have the access-date apply? Because those identifiers are permanent (commonly online copies of journal articles), and dated, it is not necessary to provide an access-date. Sources linked by |url= are, in contrast, ephemeral so it is prudent to note the date on which the content of the linked site supports the content of a Wikipedia article so that it may aid later editors should they need to locate an archived copy of the source.

—Trappist the monk (talk) 19:20, 9 June 2016 (UTC)

But doesn't the access date apply to the document itself, regardless of which of a number of identifiers are pointing to it? I have no experience with any of that, but doesn't the DOI only store the metadata + a current location, and not the document itself? (And that impression is affirmed by what I read in the lead of Digital object identifier). If the document is updated/amended, the DOI should stay the same (no?), but how would we identify which version we've looked at? Uanfala (talk) 22:00, 9 June 2016 (UTC)

Access date applies to ephemeral urls. The urls created by the cs1|2 templates for doi and the other identifiers are not ephemeral and are indexes into databases of documents. The database contains the url that the publisher of the document specifies. The publisher is free to locate the document anywhere it wants; free to occasionally move the document from its current location to another. As long as the database knows where to look for the document, a cs1|2 |doi= identifier link will always get to the document (or more commonly a paywall).

Documents that are assigned an identifier are almost never directly revised. As I understand it, when a error or some such exists that requires a change, the publisher will publish an erratum, often with its own identifier that corrects the original document.

Some authors will host a copy of a journal article that they've written on their own website. Sometimes it is an exact copy of the same article that the publisher has behind a paywall. Sometimes its a preprint. These kinds of articles are linked to with |url= at which point an access date is appropriate. Often a doi or other identifier links to the document behind the paywall in the same template as the url to the author's copy. With the access date and the url I have found archived copies on Internet Archive and so have been able to restore a non-working citation.

The only identifier that I am aware of that has versioning is arXiv. Because documents stored there are preprints, there are sometimes two or three versions of those documents. See {{cite arXiv}}.

—Trappist the monk (talk) 23:30, 9 June 2016 (UTC)

Thank you for the detailed response that showed the issue to be a bit more complex than I had thought. I wasn't aware that |access-date= is there mostly to help with locating urls. But I'm still wondering if there's any way to handle article versions. Of course, once published, articles often don't get changed, but sometimes they do, and the change isn't always in the form of separate corrigenda. (A minute of googling reveals for example this NISO recommendation for article versions, which allows for multiple versions of corrected or enhanced papers.)

Now, provided a single DOI may point to an article that has different versions (with the assumption that a new version doesn't get assigned a new DOI), then we will eventually want to have a way of identifying these versions. When in the future we come up with a way of doing that we would also want to go back retrospectively and add the version information into our citations. If these citations have the |access-date= parameter, then this should be possible. Otherwise, it won't, as far as I can see. Uanfala (talk) 12:30, 15 June 2016 (UTC)