On Wed, 12 May 2021 at 12:01, Michal Suchánek <msuchanek@xxxxxxx> wrote: > > On Wed, May 12, 2021 at 11:47:15AM -0500, Varun Varada wrote: > > On Wed, 12 May 2021 at 01:47, Michal Suchánek <msuchanek@xxxxxxx> wrote: > > > > > > On Wed, May 12, 2021 at 12:13:08AM -0500, Felipe Contreras wrote: > > > > Michal Suchánek wrote: > > > > > On Tue, May 11, 2021 at 10:43:38PM -0500, Felipe Contreras wrote: > > > > > > It is a paradox called "the bikeshedding effect". When you contribute a > > > > > > complex and convoluted change it's easier to get it in because few people > > > > > > can object (as few people can understand it). But when you contribute a > > > > > > change as simple as changing the color of something, then *everyone* can > > > > > > opine (literally). > > > > > > > > > > You forget that what you are doing right now is bikeshedding after the > > > > > fact. > > > > > > > > Except that's not what I'm doing. > > > > > > > > > You can use 'affect' or 'impact' and it generally conveys the same > > > > > meaning. > > > > > > > > That's clearly *your* opinion, but that's not my opinon. > > > > > > > > I'm not arguing between blue and red; I'm arguing between water-based and > > > > lead-based paint. > > > > > > No, you are not. There is no clear problem with 'impact', either. > > > > > > So if somebody comes along later and says that they find 'affect' > > > confusing and impact should be used does that need to be accepted as > > > well, back and forth ad nauseam? > > > > This is whataboutism and hypothetical. But even if one were to > > disregard those facts, I'm willing to bet actual money that no one (at > > least anyone with access to a dictionary or even a basic grasp of the > > English language) would do this because "affect" has a universal > > definition and is not in the realm of jargon in any dictionary or > > style guide. The same cannot be said about "impact". > > > > > > > > > The difference may not matter to you, but it matters to me. > > > > > > > > If it's bikeshedding to you, and it "gnerally conveys the same meaning", > > > > why are you arguing against? > > > > > > So if 'for' loops and 'while' loops generally convey the same meaning > > > should we accept patches that replace some 'for' loops with 'while' > > > lopps or vice versa? > > > > > > Surely not. There are different situations in which loops can be used, > > > and different people find 'for' and 'while' loops clearer and and easier > > > to understand in different situations. If you rewrite the piece of code > > > that includes a loop it might be worthwhile to change the loop type for > > > clarity, and at the time when the code is added or modified it is time > > > to discuss which one is better, not after. > > > > > > On the other hand if you state the goal to not have redundant semicolons > > > then even if code with and without redundant semicolons is the same and > > > in most cases it does not make any difference for human understanding > > > either patches that just remove redundant semicolons work towards a > > > specific goal. That makes them acceptable even if they are very minor > > > because there is clear metric they improve which makes the inverse patch > > > not acceptable. > > > > > > If you want to make the case for 'impact' in general being obscure or > > > hard to understand you will have hard time doing so. There are > > > dictionaries that recognize 'impact' as synonymous to 'affect' without > > > any difference in degree. In the COCA corpus there is around 200k > > > instances of 'effect', around 100k instances of 'affect', and around > > > 100k instances of 'impact' which makes effect/affect about 3 times more > > > frequent than 'impact'. That's not even an order of magnitude - clearly > > > not enough to claim it obscure. All of the words are within first 1k so > > > arguably if you have intermediate knowledge of (US) English you should > > > be familiar with all three. > > > > > > However, there is a different corpus that is much more relevant for the > > > git project: > > > > > > ✔ ~/git [master|…9] > > > 06:35 $ git grep affect | wc -l > > > 368 > > > ✔ ~/git [master|…9] > > > 06:41 $ git grep effect | wc -l > > > 350 > > > ✔ ~/git [master|…9] > > > 06:42 $ git grep impact | wc -l > > > 54 > > > > > > There are only 54 instances of the word 'impact' in the git repository > > > which make up only 7.5%. It is feasible to eliminate those 54 instances > > > completely. In doing so you will make the git project use the same > > > wording consistently which makes it arguably more approachable to > > > non-native speakers with limited vocabulary. That states a clear metric > > > that is improved by such patch which also makes the reverse patch not > > > acceptable and prevents potential for infinite back-and-forth changing > > > from one synonym to the other. > > > > > > Bonus points if you add a test that prevents adding new instances of > > > 'impact' in the future. > > > > So you're saying you're OK with getting rid of all instances of > > "impact"? I'm for this, but insofar as I searched the code base, I > > only found the ones I'm changing in my patch (save for a couple that, > > as a previous reviewer mentioned, are included from other repos, so I > > left those). > > Yes, I am not opposed to the change in principle. You just failed to > provide any valid reason. > > Part of writing a patch is coming up with sound reasoning why the change > is desirable and stating that clearly in the commit message. > > I don't know if this reasoning is acceptable to git maintainers but at > least there is some real data it is based on. It's useful to think of this commit via these perspectives: 1. Do you think "affect" and "impact" are synonymous? Fine; this change doesn't affect (no pun intended) you. 2. Do you think "impact" is incorrectly used here? Great, because that's what the commit is for. 3. Do you think neither of the above are relevant, but want to get rid of the word "impact" completely from the Git repository specifically because of its relatively seldom use? That's great as well, since this commit conveniently also addresses all instances of the word within the code base that are stable/controlled by the Git repo (i.e., not directly imported from other code bases). The advantage, fortunately, is that you can like any or all of these reasons, and we don't have to agree on which ones are the most important or relevant. The end result is the same: a less ambiguous code base that makes everyone happy.
