On Sun, 11 Feb 2024 21:07:42 +0100 Thomas Goirand <zigo@debian.org> wrote:
Back in the days, I would have understood why things would have been written that way: the Oslo libs were young, and not available everywhere, and it was kind of a good idea not to use them.
So I am wondering: - is it still the case, that I shouldn't use oslo.config / oslo.log when adding a patch in Swift? - can we revisit this, and convert swift to oslo.config generated config files, and PBR shell-script entry points instead of binary simply dropped into the toplevel bin folder?
I think Swift being different here is a technical baggage at this point. Switching over will please some, but upset the constituency of heavy users of Swift. We can try to stay compatible, of course. At my side, the get_logger() was particularly problematic when keystone token middleware was shared, because it uses LOG and that just does not produce any output. I never had an issue with logging_monkey_patch(), but I can see that too. The main problem is just that Oslo drags in a great many dependencies, and Swift used to be installed standalone by many operators. With container images becoming the prevalent installation method, this concern may be not as significant as it used to be. But honestly I would prefer to register your concern and leave it be. Too much other work. BTW, I'll take a look at that drive checker review. -- Pete