[R] Deprecating download method='wininet' in R on Windows causes trouble with corporate proxy

[Henrik Bengtsson]

Is R centrally installed?  If so, environment variables 'HTTP_PROXY',
'HTTPS_PROXY', and 'HTTPS_PROXY_USER' could be set for all users by
setting them in the R_HOME/etc/Renviron.site file.  R_HOME is the
folder where R is installed.  You can find this file from within R by
calling:

> file.path(R.home("etc"), "Renviron")
[1] "C:/PROGRA~1/R/R-42~1.1/etc/Renviron"

If not centrally installed, I don't know anything better than users
setting them in their personal ~/.Renviron file;

> normalizePath("~/.Renviron")
[1] "C:\\Users\\alice\\Documents\\.Renviron"

For example,

> cat(file = "~/.Renviron", append = TRUE, "HTTP_PROXY=http://proxy-host:3128/", "HTTPS_PROXY=https://proxy-host:3128/", "HTTPS_PROXY_USER=dummy", sep = "\n")

At least this avoid having to configure them in MS Windows settings,
which is tedious to document and explain.

My $.02

Henrik

On Thu, Sep 29, 2022 at 3:48 PM Selke, Gisbert W.
<gisbert.selke using wido.bv.aok.de> wrote:
>
> Method="wininet" is deprecated and scheduled to go away, the standard method is now libcurl. This causes trouble for all R users in our shop, because we are sitting behind a corporate proxy, which uses Kerberos authentication. (We're all on Windows.)
>
> Using wininet, this used to work without problems and without additional effort; it currently still does with explicit method="wininet" (which, by the way, precludes use of the handy menu command "Update packages", which will use the default method, i.e., libcurl as of now.)
>
> For the future, when wininet will be gone for good, the only option we have is to resort to first setting environment variables HTTP_PROXY and HTTPS_PROXY and then tricking the proxy out of using Kerberos, setting HTTPS_PROXY_USER to a dummy string.
> This is certainly doable for R users with enough knowledge of the technicalities of internet access, but our average R user will just be lost. As has been pointed out elsewhere (https://github.com/rstudio/rstudio/issues/10163#issuecomment-1154071514) , this will create a lot of blood, sweat and tears (and swears), and it is a moderate nightmare to maintain consistently and up-to-date for many users.
>
> My first question is: Since we are probably not the only institution in this situation, has anyone come up with a robust and maintainable solution other than our approach described above?
>
> Failing that: would it be possible at all to change the use that the R core makes of libcurl in such a way that it would automagically Do The Right Thing (tm)? In principle, this should be possible; after all, wininet did the trick, and ordinary browsers can handle this situation. (Disclaimer: I know nothing about the R internals so cannot say whether I am being overly naïve here.)
>
> Any help appreciated.
>
> (I'm new to this list, so if this has been discussed here before, I apologize and would be grateful for a pointer to do my reading.)
>
>         [[alternative HTML version deleted]]
>
> ______________________________________________
> R-help using r-project.org mailing list -- To UNSUBSCRIBE and more, see
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
> and provide commented, minimal, self-contained, reproducible code.