Christian Grothoff transcribed 3.7K bytes: > On 02/24/2018 10:43 PM, ng0 wrote: > > Marius Bakke transcribed 1.6K bytes: > >> Tobias Geerinckx-Rice writes: > >> > >>> nckx pushed a commit to branch core-updates > >>> in repository guix. > >>> + ("nghttp2" ,nghttp2 "lib") > >> As far as I can tell, we don't have an nghttp2 package yet :-) > > > > > > Why? I don't even intend to support http/2 with gnURL. > > gnURL is not cURL. Are there any reasons why this should be build in? > > Well, the idea of Gnurl was to be reasonably minimal and to avoid the > kitchen-sink dependencies (and unpredictability wrt SSL support) of the > original Curl. I personally don't see a need for http/2 at this point > (bad), especially as it adds additional dependencies (very bad) and > would not be available equally on all systems Gnurl runs on (very very bad). I agree here and wasn't 100% sure on http/2. I just check the list of abilities to not build only when I do the releases. > My recommendation remains that Curl should be modified to dlopen() > extensions/plugins to load protocol/TLS-support extensions on-demand. > Until that happens, This will take some time, cURL has some of your early criticism on their todo list last time I've read the todo list, but we might be faster switching to wget2. Only time will tell. > my view is that we should be very, very conservative > in terms of features we add to avoid creating another kitchen-sink. For > those with kitchen-sink requirements, there is always Curl... Okay. With this in mind, could anyone with access to the branches in Guix revert this commit? Furthermore, should I add a comment above the gnurl package definition wrt the goals of gnURL and that it already gets build the way it is intended to? > _______________________________________________ > GNUnet-developers mailing list > GNUnet-developers@gnu.org > https://lists.gnu.org/mailman/listinfo/gnunet-developers -- ng0 :: https://n0.is | https://crash.cx A88C8ADD129828D7EAC02E52E22F9BBFEE348588