Re: Building libcurl for multiple Linuxes / CA store
Scott Talbert via curl-library <[email protected]>
| Newsgroups | gmane.comp.web.curl.library |
|---|---|
| Message-ID | <[email protected]> |
On Sat, 7 Dec 2024, Daniel Stenberg wrote: > On Sat, 7 Dec 2024, Scott Talbert via curl-library wrote: > >> 1) Extend the ca-embed functionality so that it works with libcurl and the >> CA store could be bundled with libcurl. Is there a reason ca-embed was >> restricted to just the curl tool? > > I argued for the logic to be done in the tool, because I don't think it is > necessary for the library to do it. A bundled CA store is complicated already > as it forces you to update the application regularly instead of just updating > an external text file. > > After all, it is ultimately the user of the library that decides what CA > store to use and the application is the user. > > If you want this functionality for your application, you can just do the same > as the curl tool does. libcurl provides the APIs that make it possible. > >> 2) Implement some sort of runtime CA bundle auto-detection, similar to the >> compile-time one. It seems there is already support for this on Windows, >> but similarly only with the curl tool. > > Why do this in libcurl when you can with less effort do that logic in your > application? Then you can make it work exactly the way you want it to, > without having to care about others. Calling pycurl an application is a bit of a stretch, but I suppose from libcurl's perspective, everything is an application. :) Initially, I wasn't thinking it would be possible to do runtime CA bundle auto-detection in a way that's transparent to end users of pycurl, but after looking at how libcurl handles CURL_CA_BUNDLE (it seems to ultimately do what setting CURLOPT_CAINFO does), I think it should be fine for pycurl to set CURLOPT_CAINFO before handing over the handle to the end user. Thanks, Scott -- Unsubscribe: https://lists.haxx.se/mailman/listinfo/curl-library Etiquette: https://curl.se/mail/etiquette.html