[R-pkg-devel] DOI in Description (possibly) invalid

[Ben Bolker]

   I wonder if it would be possible for CRAN to use a DNS caching proxy 
(something like Squid) for DOI requests? It seems like that solve most 
of the problems (requests would only happen when a brand-new DOI was 
encountered, most requests would be for DOIs that had been looked up 
previously ...)

On 2022-05-08 7:29 a.m., Ivan Krylov wrote:
> On Sun, 8 May 2022 08:01:16 +0000
> Murray Efford <murray.efford using otago.ac.nz> wrote:
> 
>> This results in checks reporting "Found the following (possibly)
>> invalid DOIs:... Service unavailable".
> 
> This happens because there's Cloudflare sitting between the world and
> Wiley Online Library, and Cloudflare replies with the HTTP error code
> 503 to requests it doesn't like. When checking DOIs for validity, R CMD
> check doesn't bother pretending to be a human operating a browser (it's
> hard and keeps getting harder), so Cloudflare classifies this request
> as automated and possibly malicious and refuses to serve it. On the
> other hand, when you follow the link manually with a browser,
> Cloudflare knows you're a human because it's been keeping track of your
> browsing patterns across other websites under Cloudflare protection,
> and they match typical human behaviour, so you get the content.
> 
> It's a known problem. For now, you just have to mention it in the
> submission comment.
> 
> There has been a suggestion to change the DOI check so that it would
> stop after one request to the https://doi.org/ server (which doesn't
> discriminate against such automated requests for now) and doesn't try
> to follow the link returned by doi.org (which might). Perhaps it would
> get implemented when this problem gets worse.
> 

-- 
Dr. Benjamin Bolker
Professor, Mathematics & Statistics and Biology, McMaster University
Director, School of Computational Science and Engineering
(Acting) Graduate chair, Mathematics & Statistics