Hello, Fabian. Thank you for detailed explanation.
On 2023/05/15 16:35, Fabian Thorns via lpi-examdev wrote: > The Linux kernel is different. It has established QA mechanisms, > contributors will receive feedback and will get acknowledgement and > recognition when their patches make it into the kernel. A certificate won't > change anything here, it wouldn't make any changes be approved earlier, and > a list of actual commits is likely worth (at least) as much then a sheet of > paper. First, there is a barrier (prerequisite knowledge that does not depend on a specific domain) before a new developer can submit a patch. Second, is QA working? When someone submitted a patch, correctness of idea is reviewed, but correctness of implementation is hardly reviewed. Checking correctness of implementation is up to open-testing. Many developers think that it is OK if no problem is found. Developers might think that that is OK because there is no warranty. However, for me who had been in security field (or rather, secure coding field), I don't think that we are successful in sharing enough skills/knowledge for correct implementation. Since the kernel depends on various areas in a complex manner, developers cannot do their works only with knowledge of their own area. Due to limited reviewers in their own area, unexpected problems are occurring after the patch was merged. Some mistakes are trivial to fix, but others are complex problem which involve multiple areas and subsystems. Problems which involve hierarchies/dependencies, especially locks, are difficult for an outsider to understand. Even veteran developers may not know the specifications of other domains on which their domain depends. I am doing bug fixing activities that are not limited to a specific area. Basically everything is in other areas, and there are many things I don't know. Therefore, I want maintained knowledge/skills/examples which are written at a level that any kernel developer (including newcomers and outsiders) can understand. This is the background of why I want this certification exam. We forever need specialists who have a detailed understanding of specific areas, but I would like to increase the number of generalists who have some understanding of various areas. Existing material is either scattered and hard to find, out of date, or requires too high level of prerequisites which is too difficult for newcomers. (I can't understand without example code showing how the code works.) Is it possible to use LPIC as a mechanism for both veteran and newbie developers participate and review the exam content serving as an entity that fills the gap between veteran developers and newbie developers ? Since people can find answers by searching the Internet, exams that ask examinee to answer command names and option names/values are not useful to me. I want this certification exam to ask examinee whether the concept and content are understood based on the latest information/knowledge, by requiring periodic updates. The exam content of the questions will become longer than now by citing images and source code fragments. But since LPIC is a CBT exam, it will be possible to cope with the increase in the amount of scrolling on the screen. If this certification exam can be realized, it will be useful for fostering kernel developers and improving kernel quality. I explained my expectations from the viewpoint of quality improvement at the kernel source code level, but for example other contents such as what projects related to the kernel exist, how to use tools that are available for kernel development etc. can be included in the scope of this exam. > > That doesn't mean that it wouldn't be nice to have mechanisms to facilitate > the onboarding of new developers, Not limited to new developers. > but those would more likely be education > or mentoring programs -- eventually, a certification is at the end of an > education journey, not at the beginning. This certification is meant to be at the beginning. An indication of having basic common knowledge for kernel development. > Similarly, keeping up to date with > changes in APIs and project structures will likely be better done with a > newsletter, podcast or something similar. Problem of Linux kernel development is that not all developers are aware of changes, for Linux kernel project is too big to be aware all. > Tetsuo mentioned sharing > information via a certification, in this case the actual benefit would be > exam objectives. Some examples: How to use locks correctly. How to use memory management subsystem correctly. How to debug problems (kernel oops / warnings). > If such a knowledge sort would happen in a medium that > allows more meat on the bones, the benefit will likely be higher. I am not > sure what resources exist to get kernel developers started and keep them > updated, if anyone has some good resourced to get started, feel free to > share :)) Regards. _______________________________________________ lpi-examdev mailing list [email protected] https://list.lpi.org/mailman/listinfo/lpi-examdev
