On Wed, 15 Sep 2021, Isaac Jurado via curl-library wrote:
I was curious why optimizing the storage is so important when most of the cookie handling should be done without storing them on disk at all...
From what I understand, right now it's not possible to enable CURLOPT_FOLLOWLOCATION and be able to manage cookies from "user code" at the same time.
That's true. If you want to manage cookies yourself, you should probably just disable CURLOPT_FOLLOWLOCATION and handle redirects yourself as well.
Looks like a trade off: if I want to handle cookies but not with libcurl's cookie jar, I have to deal with 3xx responses myself too.
The cookie jar (the external file) is only populated by libcurl when you close the easy handle. Mostly you re-use the handles with the cookies in memory.
To give more context, I have been prototyping an utility to perform authenticated service calls over HTTP. Authentication is OAuth based, but I want to maximize session cookie reuse. Therefore, an SQLite based cookie store fits very well even in a multiprocess environment (not only multithread) thanks to the ACID transactions.
Isn't using a share handle for cookies the better solution for this? -- / daniel.haxx.se | Commercial curl support up to 24x7 is available! | Private help, bug fixes, support, ports, new features | https://curl.se/support.html -- Unsubscribe: https://lists.haxx.se/listinfo/curl-library Etiquette: https://curl.haxx.se/mail/etiquette.html
