As programmers I feel like we'll always nitpick and bitch over what the optimal setup is for rather mundane things (tabs v spaces, yada yada).
I'm not saying that conventional commits are God's given best way to structure a commit message, but they are a defined structure, and I find it much more effective and important that some expectations be set around commit messages, and I think conventional commits are as good as anything.
Like the author is making a big deal that they think scope is more important than type. I may tend to agree, but I think the difference between "fix(compiler)" and "compiler fix" is not exactly a hill I'd be willing to die on.
The tech industry has tons of things that became standards even if they weren't optimal. E.g. if one were starting from scratch I think any sane person would argue JSON should support comments (sorry but Douglas Crawford's rationale for not including comments never made sense to me), better defined numeric formats, etc. But it was better in many contexts than what came before it, so it became the standard. I could believe that there is some other format that differs a bit from conventional commits that is a little better, but not really better enough to want a whole other competing way of structuring comments.
Defined structure does not constitute quality. A commit message can be loosely structured, but be very insightful and good at communicating the nature of the change. On the flip side, one can make a very structured but confusing or non-informative commit message.
I generally tend to agree with the author, conventional commits do not solve the core issue of the poor commit messages problem.
Conventional commits made it easier to generate changelogs and automate semantic version bumps. I suppose LLMs can usually do that the right way with looser structure, but in the before times it made a lot more sense, and even now is much less ambiguous
It’s the linting argument all over again. I don’t give a shit what the style is. I give a shit that it’s consistent. Form whatever opinion you want about how you want to format your code, structure commit messages etc. I don’t really care, if you want to start every commit with “poop(fix): pooper my commit message”, as long as you’re consistent about it and enforce it programmatically you have my emphatic support
This comment assumes you are picking between a conventional commit message or something better. But the reality is you are almost always picking between a conventional commit and Nothing.
That is only true of semantically equivalent things.
Tabs vs spaces don't matter they are equivalent.
But consistency is only better when it is an improvement.
It is fundamentally important that convential commit is better for adopting it to be an improvement.
After all in your example wasting the first four characters of your commits with poop would objectively reduce the quality of your commit history, whether or not it was consistent.
> Tabs vs spaces don't matter they are equivalent.
Just to nitpick (because what else is this thread about? :))
They aren't equivalent! Tabs carry more semantic information than spaces. 1 Tab character == 1 Level of nesting
Space-based systems _can_ provide the equivalent semantic information if they are 100% consistent.
...but part of the argument in favor of spaces is that they allow an escape hatch of the strict indentation in order to allow pleasing visual alignments.
why is "solving" the issue somehow the bar? software engineering has more practices rooted in psychology than engineering, its a moving and ambiguous target. Using conventional commits gives you a framework and mechanism that undoubtedly improves contribution semantics.
In what context does wasting your first characters on fix vs feat matter?
PRs are going to have an explanation that has way more detail than necessary to figure that out quickly.
One lines tend to be (for me) in a situation where the difference is immaterial. If I am rapid firing through history I need to know what you did not why you did it.
Again I am not claiming that these are bad or even that they aren't good.
I am specifically disagreeing that any change is automatically good, that isn't true.
> Using conventional commits gives you a framework and mechanism that undoubtedly improves contribution semantics.
I do not want to contribute to a project using conventional commits. I have consistently found, that I am unable to decide what the "type" of a commit even is and I feel unnaturally caged in into how I would split up commits, by having them be restricted to types (it doesn't help that the conventional commits I've seen appear to decide the type by fair dice roll).
Discouraging contributions does not "undoubtedly improve contribution semantics".
Great perspective! As a sidenote, I really like json5 for everything that would be needed for a human to read and it has great support everywhere. It fixes so many of JSONs shortcomings like comments.
I’m all for standardization but you could just use this argument to keep any suboptimal status quo in place. XML is good enough and a standard. SOAP is good enough and a standard. etc.
The claim is that Conventional Commits are good enough and standardized enough that having another structure isn’t really worth it. But “worth it” is subjective. I’d say that if you are making commits and reading PRs every work day, and the conventional commits format causes a little bit of friction, that friction can add up. Having another option other than seeing conventional commits as a law of nature gives options for teams who prefer it. (Most teams aren’t generating changelogs anyway.)
> Like the author is making a big deal that they think scope is more important than type. I may tend to agree, but I think the difference between "fix(compiler)" and "compiler fix" is not exactly a hill I'd be willing to die on.
The big deal is not that scope is more important then type, the big deal is that natural language allows you to formulate things to emphasize whatever you consider important, and by forcing everything into a specific format this information is lost. There's a reason we have formats like markdown and plaintext, not just JSON.
I don’t have a horse in the race, but the OP objections ring hollow to me.
- scope is important: true, but isn’t that derivable from the commit contents? An important sanity check on a diff is to look at the paths touched. (A “test” diff should not modify prod auth code.) but if you want to see this from —oneline, sure, I think feat(auth): is better than feat:
- wrong audience: I don’t agree. feat commits _should_ actually describe the product-facing changes. You _should_ curate a nice stack with your no-op refactor changes first, then your small new feature change atop. This is the most helpful thing to include in a diff comment. You should put anything technical in comments so they are not lost, “why I chose algorithm X” belongs in a comment or DECISIONS.md. These are all tedious things that only psychos bother doing in a commit history at a fast moving company, but on OSS projects I think it’s much more important to stash context in the commit messages.
I end up thinking that conventional commits is a good idea because you can make tooling force people to at least put one iota of thought into what goes in a commit message. I've had to review so many commits with the subject "small fix" that was not in fact a small fix.
I'm not yolkedgeek but I can give my own answer: EDN has tags. Tags start with `#` and are followed by a symbol (which is a lot like an identifier except that a lot of punctuation is allowed in symbols, because EDN derives from Lisp syntax rules). The `/` character is used for namespacing, and a user-defined tag must use a namespace. The tag meaning is application-defined, but there are a couple standard tags with well-defined meanings:
#inst "1985-04-12T23:20:50.52Z" = an instant / timestamp in RFC 3339 format
#uuid "f81d4fae-7dec-11d0-a765-00a0c91e6bf6" = a GUID/UUID
More tags could be defined by the standard later, because the entire unprefixed namespace is reserved. But just having a well-defined way to represent timestamps and UUIDs is an immense win over JSON, where you have to somehow know (based on what you were expecting to receive) that this string should be parsed as a timestamp or a GUID.
Also, user-defined tags will often be used to represent a class:
#myapp/Person {:first "Fred" :last "Mertz"}
Again, no need to know (based on what you were expecting) that this particular object is an instance of Person; the data transfer format tells you what class it is. JSON has to add a field, and what field it is will vary from application to application so it's usually not possible to write a universal parser. One server might generate { "__type": "Person", "first": "Fred", "last": "Mertz } while another one does { "$$class": "Person", "first": "Fred", "last": "Mertz }, for example.
EDN also has syntax defined for sets, but that's a smaller win over JSON, because it's not often necessary to declare that something is a set. Still, there are times it's helpful; it's certainly not a bad thing to have a set syntax.
Also, EDN has comments built in to the system. Two kinds, one line-based comment (useful for actual comments, e.g. when you use EDN as a config format), and one that comments out the next thing in the file (useful for temporarily commenting out an entire section with a single token, or for removing ONE item temporarily from a list that's all on the same line so line-based comments are difficult). Because Douglas Crockford didn't envision JSON as being used for config, he forbade comments in JSON, and people have been coming up with competing proposals for putting comments back in ever since. (Thankfully, nearly all the proposals interoperate, because all of them sensibly use Javascript comment syntax, so it doesn't matter if the file is JSONC or HuJSON or JSON5, the comment syntax is the same).
But the biggest win for EDN is tags, which can convey type information outside the data structure. JSON has to use something inside the data structure to convey type information, and there's always that small chance that the name chosen (__type or $$class or whatever) will collide with a property of the actual object that was supposed to be serialized.
I get tags and atoms. It seems like the problem with class serialization is somewhat arbitrary though. It seems like both sides need the object schema ahead of time, in which case the schema can flag how it sdould be IDd / tagged.
I also wonder if atoms can be reduced for low-bandwidth transmission. Naïvly, you could just prepend a lookup table for multiple-use atoms.
I guess it seems more like niche, additional features when GGP seemed to be claiming a big step up.
hence tabs. just configure your display for the width you like.
though I do wish more editors had some kind of "ignore the spacing, display it semantically like [this]" and just let you insert whatever you wanted, converted to whatever is nearby, and didn't touch lines you didn't create. there's no reason to even have the debate or care about inconsistencies, you can essentially always* convert between them losslessly in terms of behavior.
The next iteration is "smart tabs", where indents are stored as spaces in files for consistency, but when using an editor it recognizes that something special is going on, and offers you a tab-like experience.
which they kinda already do, yea. I just want it to be stronger, and to correct indentation that doesn't follow your formatting rules (which I have not seen / is not widespread). kinda like a `gofmt` overlay on a file, rather than modifying the file directly.
as a bonus, this could also make `goto fail;` errors more obvious.
I don't need a standard for this. This is just noise. There are some people who have some sort of mental ailment that makes them obsessively want to introduce "structure" "scheme" "patterns" where it is just innately nonsense. You do you, but stop trying to force it on people.
Reminds me of the "scrum master" adjacent folks who could never cut it writing code and then branched into all kinds of things like "Git Flow" when having never understood Git to begin with. Peak bikeshed territory.
Reminds me of a place I worked at where a "naming committee" had to approve variable names. And no, you could not use "i" as an index in a one-line loop.
I work with legacy code and am regularly reading commit messages from 10-15 years ago while figuring out what was going on several teams past. It's also why I'm against squash-merges, there have been plenty of times a commit was in the middle of a chain of commits and the helpful context would have been erased if we had squashes in svn.
The real takeaway is that different projects have different requirements.
In over 30 years of using source control, I've never once worked on something where it's useful to include the component (article calls it scope) in the description in a standardised way. It's obvious what components are affected based on where in the source tree the affected files are. Similarly "bug", "fix" or "feature" adds no useful value. It's important or it wouldn't be checked in.
The only thing I've found useful, and which the article doesn't even consider, is a link / id for the relevant change request. The commit already contains all the information about what was done in the change, what's missing is the context about why.
Even on my solo projects I include a JIRA reference in square brackets before the description. If it's just something I randomly decided to fix during the course of development, I'll create a short 1 line JIRA to get an id and explain the why there.
> The only thing I've found useful, and which the article doesn't even consider, is a link / id for the relevant change request. The commit already contains all the information about what was done in the change, what's missing is the context about why.
The "why" is THE thing that needs to go in the git commit message. Capturing "why" is the entire point of that message and slapping a link to some external (and eventually absent) resource is not a good substitute.
Respectfully, I disagree. A good commit message to me is something like:
[PRJ-123] Changed blah to foo
Blah didn't handle the wangle flange properly in some cases, foo is a better fit for customer requirements.
The "why" that justifies the change, is already contained in the JIRA ticket PRJ-123 and explains exactly what the customer requirement was that necessitated the change. It will almost certainly contain a lot of detail that isn't relevant to the commit message, because that isn't the place to be documenting customer requirements, and probably relates to a number of other tickets. Perhaps the code itself might have a comment explaining the code change, if it's a non-obvious implementation, but otherwise the ticket is the best place for that information.
Additionally, if a change requires multiple commits, you don't want to be repeating the justifications for the entire feature in every commit message. It's redundant. But the commits will all be tied together by the ticket reference in the commit message.
The difference is that the Jira ticket is for everyone involved in a project (business analyst, UI designer, QA, support, DBA), while commit messages are written with developers being almost exclusive audience. PRJ-123 might explain why an end user might need it, but the commit message explains why the change (diff) is the way that it is. The ticket answers requirements-level questions, the commit message answers code-level questions. Commit messages are useful both during the review and when a future maintainer is reading the code.
> Additionally, if a change requires multiple commits, you don't want to be repeating the justifications for the entire feature in every commit message. It's redundant. But the commits will all be tied together by the ticket reference in the commit message.
Different commits do different things, so require different justifications. Here's a fictional example to demonstrate:
First commit:
[PRJ-123] Server: extract class Foo
In the next commit, we're going to need to re-use the foo logic from
class Bar. Extract new class Foo from Bar to make it available for
re-use.
Second commit on the same ticket
[PRJ-123] Server: use Foo in Baz
The users of BazClient need to be able to see foo information in the
baz dialog. Include Foo in the data sent by class Baz in the server.
Side note: the user might not even know that they are looking at Foo and Baz, it might be called something else in the UI they are shown. Whether or not this needs to be included in the commit message depends on the situation.
And later in a commit fixing a bug:
[PRJ-456] Server: check ID for null in Foo
When class Foo was extracted from Bar in commit deadbeef ([PRJ-123]
Server: extract class Foo, 2026-06-06), a null check for the field
ID got lost by accident.
Check the field ID for null in class Foo to avoid a
NullPointerException when a foo event is sent to Baz.
Hah, maybe its the difference between why as in "this addresses problem X" and why as in "because those jerks in pre-sales sold another imaginary product"
I think the commit ought to describe the purpose of the change in terms of its result for the software's intended use. Feel free to hide the business/political drama behind a ticket number.
This gets down to a more fundamental tension. Are commit messages to communicate between developers? To communicate from developer to consumer? Or some kind of project manager golem? In practice, it is usually some constantly wandering attempt to be a blend of these.
The last part is easy to answer. Commit messages are solely for developers IMHO. The communication between developer and customer / product manager should be via the ticket system.
That said, knowing the commit ID something is fixed in, so that the PM can track what build it emerges in is useful.
I find bug trackers and source control fad change several times over the life of my code. A number from a ticket system we no longer use is not helpful.
But a number from a ticket system you are using is helpful and vastly more log messages will be read during the time when it’s active than after it’s been retired/replaced.
1. Usually the commit message is often too short to capture the "why" adequately.
2. It is very beneficial to capture the why in one single source of truth, and that usually is not the Git commit message in a business context. Hate on Jira all you want, but if you capture the "why" there, you can add comments, view history, add rich context, link dependencies, add rich context, etc. Can't do that in a Git commit message.
My ire is more directed at github PRs than Jira... but the same basic idea applies. You want a single source of truth and you want that as close to the origin (the code) as possible. Your history, dependencies, etc. are all in git already and can be highlighted there if appropriate. For general comments, git notes covers that.
Business (ie. $work) will dictate whatever it wants and that is what get used but for anything I personally have control over, everything goes in the repo itself to prevent platform lock-in. For example, github's been going downhilll lately but all those projects with their history in PRs, etc. now needs to exfiltrate all that data somehow.
For GH issues you can always navigate back to the PR discussion (which should have linked issues and other pointers in it) from the commit.
Of course when we switched to GH issues, we largely abandoned JIRA and years later the instance got turned off and deleted. Now all those JIRA tags are entirely useless.
IMO that actually argues for tight coupling between your issue tracker and your git repo. And what you really want is portability (which I don't see how you get other than with tight coupling). Ideally there would be open standardized formats, but as it is, github is the 800# gorilla that defines the format and as long as gitlab and other clones can slurp in github project metadata (or at least PRs) then that effectively gets you closer.
But any way... Fixed, immutable pointers to an Atlassian product that you might not be using in 5 years is not a good policy. I'd sooner accept the policy that the git commits needed to stand entirely on their own and all the information about the "why" of the change needed to be baked into the git commit or the comments in the source (I think that fails, though, since everyone is overly terse in git commits and summarizes issues and loses information--and the back-and-forth dialog in a PR discussion is useful because it contains more than just one person's voice summarizing the reason for the change).
> Of course when we switched to GH issues, we largely abandoned JIRA and years later the instance got turned off and deleted. Now all those JIRA tags are entirely useless.
We did Bugzilla -> FogBugz -> Jira. Almost all the data was lost every time, no one bothered with migration except for the maintenance project. Worse, even on Jira we lose cases as teams end and hand off the code, and the Jira project is closed so no one else can access it.
We've also done cvs -> svn -> git. All the commits have survived migration.
I do keep including cases in the commits messages, if nothing else it'll help link things together in the future, but never rely on them for context a future maintainer might need.
> we largely abandoned JIRA and years later the instance got turned off and deleted
Sorry to be nitpicky, but why did you abandon a tool that contained a lot of valuable knowledge? That's not the fault of GH nor JIRA, that's your fault. At least you'd back up descriptions + comments from these JIRA sources.
I didn't abandon it, it was run by a different team, and we were one consumer of it. When the organization switched from Jira to GH issues the Jira was kept running for years, but nobody got the information into GH issues. Eventually the Jira was shut down, certainly by the time the company got acquired.
That isn't true though. It's very easy to export your data from JIRA. From your board, go to the List tab, filter the items to whatever you want, and then click ... and you can export the data in various different formats. Exporting as XML dumps everything.
You can trivially export your data from JIRA. If the parent experienced a situation where valuable information was lost because the instance was deactivated, that's not JIRA's fault.
It is useful if you automate generating release notes. Then your notes are grouped by new features first, then bug fixes after. This makes it a little easier for non-technical uses to read.
Absolutely not. Commit messages should never be automatically passed through to the end-customer. I also worked in a place that tried it once and it was a disaster. Sure, a list of commit messages can be a useful start as a list of things that might want to be put in the release notes, but very rarely is the developer the right person to be explaining those changes to the end user.
If a developer is being asked to do that, it's a good sign that the PM isn't doing their job properly.
Exactly. The commit message is supposed to be for the future developer, not to generate changelog.
And the main case when that developer reads the commit message is when he doesn't understand _why_ that commit exists. Not what it changes, but what is the purpose of certain lines. So he runs "blame", sees commits, the original developers are not with the company anymore, the old JIRA may not exist too, and the only hint is the commit message.
This, along with the "successful git branching model", are symptoms of the fact that devs overwhelmed with the flexibility of git and look to other people to define standards for them because they that lack the experience to do so for their own requirements.
Actually, this is also similar to classic OOP, where people use a contrived method of structuring their code.
The use of the word "chore" in many users of conventional commits has always riled me. I've always tended to favour the "linux kernel"[0] style of commit subject, which thankfully gets a mention here.
Completely agree, the attitude implied by “chore” is very off-putting to me. As if the rest should all be marked “fun” or “indifferent”. That kind of emotional judgement doesn’t belong in a commit message.
I’ve never personally used the chore term, but it doesn’t bother me to see it and I don’t feel it has a negative connotation.
Cleaning my kitchen after a meal may be a chore, but it’s not an intrinsically bad or unpleasant experience most of the time, it’s just good hygiene and afterwards I have the satisfaction of things being clean. Not cleaning the kitchen feels way worse to me as it ultimately leads to other far more unpleasant situations.
Such it is with updating dependencies, it generally needs to be done, so it’s good to do it, but it’s in no way noteworthy, so chore describes it perfectly, to me it signals that: “it’s work that needed to be done, but not for a feature, functionality change or bug fix on this particular code base, so you’re unlikely to see much change”.
You just made me realize why I've always considered 'chore' the most ambiguous type. In addition to being loosely defined ("transparent change with zero functionnal impact"?), this one is indeed a word related to emotion. No wonder it has a more subjective meaning than 'fix' or 'feat'.
This is why I never use it and almost always pick 'feat' to please the linter. Because I can't help considering that any change worth committing is improving the quality of the code in one way or the other, and thus a feature.
It is bad terminology, yes. But also - a pretense that you know the overarching influence of a commit ahead of time, which you don't - but once you have conventional commits everyone on the team and the LLMs have to spend time/tokens inventing that stupid nomenclature.
My main complaint with conventional commits always was that they don't include an issue number in the commit title. It's not even mentioned in their standards as optional or something.
To me this is almost the most important information in a commit message. I don't know how often in the last 15 years I was cross checking the issue description referenced by some old commit to get the full context of a change. I also felt that this habit is kind of standard - until i had to learn about "conventional commits".
I personally prefer including issues as git trailers:
fix thing in foo
Issue: ABC-123
Git has plenty of builtins for parsing and formatting these trailers, so you can easily create custom git log aliases that let you see them inline and parse them for use with CI.
I think you’re the first person in this comment section to mention this. Git has a structure for machine-readable structured data. Put it there, not in free text fields for humans.
Personally, if I am skimming a change log that is already limited in characters, I don’t care about ‘XYZ-999999’ in the main commit message. It’s good to tag as a trailer but I’d much rather see what the commit did than the Jira issue it came from.
It's totally fine to put the issue number somewhere at the end of the commit message, and not in line 1.
Most tools cross-link them as long as #<issue-id> is mentioned anywhere in the message. It's also useful the other way around, open an issue and see all associated commits.
Yeah the ticket value falls off pretty quickly to me. If I pull that up and it's been a closed issue for years and code has been added, rewritten, people moved, and tons of other changes to where the ticket is just a historical artifact and doesn't need priority in the first line of the commit message.
IMO that is only a problem for those who demand that the issue key needs to go first in the subject (which again IMO is bad for readability). I don't see why you can't just stick it somewhere after all the conventional commits junk? Issue keys ought to be something that can be janked out based on a regex like an alphanumeric prefix followed by a number, so things like this "standard" have little need to set aside a space for it.
Personally (without conventional commits) I tend to put them at the end in parentheses if the commit has something to do with that issue. But if there is a stronger relationship like that it fixes the issue, I put a `Fixes` trailer in the message as well.
Interesting, I guess we've been doing it wrong this whole time, as we do `fix(ABC-123): some message here`, which ends up being linked great, renders very nicely into the automated release notes, etc.
It's only a convention around commits suggesting to prefix them with certain short words, there's nothing stopping anyone from including issue numbers.
All ours at work do: "feat[component]: DEV-1234: description"
yeah this is the actual key. an actually useful title and a stable link to the discussion around the change.
conventional commits are pleasing, but questionable actual utility. the code speaks for itself. the actually useful information is a well chosen title and the context for the change.
Why do you even want the issue number in the commit title? I find that super annoying and unfortunately GitHub kind of forces it on you if you use merge queues.
if IDE supports converting them into clickable links, then scrolling through a log in its Git integration (kinda like `git log --oneline`) will be a convenient list of links to issues. If the commit messages are good, you just skip over the Jira issue/ticket key/ID. If what you want is to see the requirements of a feature or steps to reproduce of a bug or some other context for the ticket, then a clickable link is very convenient. Of course, sometimes it does make sense to include some of that information in the commit message as well.
When working in big teams, it can be very hard to increase the usefulness of commit messages. On the other hand, enforcing inclusion of a Jira issue key in commit messages is easy to implement. Relying on issue tracker descriptions can be a difficult proposition as well. Quality of individual ticket descriptions can be low; depending on how responsibility for maintenance is handled, the bug tracker migrations can sometimes be handled improperly and information can be more easily lost than in a Git log.
It’s very helpful to know the motivation for the commit and if that motivation was tied to a client contract/feature. Especially in cases where a commit affects multiple files or even just one file so that all commits can be grouped into a feature/contract.
COMPANY-1234 in the title doesn't tell the reader all that much about the the feature or motivation. It does tell the client, but I'm not seeing why that is better than having it in the description as a tag, or some other nice way of extracting it.
Least of all when that ticket is older and so much of the code and the company has changed too. Like sometimes useful historical context sure but worth putting in the first line of a commit? I put it in the body with a link to the ticket or tickets as a footer, if someone wants historical context it's there.
I can not say anything nice about conventional commits. The format takes up space in the most-read part of the message. The categories or types have little information. They can be replaced with an honest English verb embedded in the subject like a sentence. It also reads way better with just a sentence instead of three kinds of punctuation (:, (), !). Okay, I can tolerate an "area" in the subject. And that predates this conventio.
At my dayjob we make a webapp for non technical people. I can write a changelog for that just fine (in norwegian). The commit messages are irrelevant to the users. And demanding that all commits should be good enough for an end-user changelog? That's not happening for us anytime soon.
Conventional Commit messages have big https://goomics.net/361 energy. Or at least the mechanisms that repositories use to programmatically enforce their use.
(I can't believe nobody has pointed this out yet.)
Inverting the order actually addresses my primary annoyance: what is a feature?!
> refactor(core): Update webmcp support to use document.modelContext
As the author points out, the line between a fix, an improvement, and general clean-up is blurry and dividing each semantic change into its own commit (and possibly squashed later anyway) is just creating work for no one's benefit.
I think Conventional Commits are just an artifact of trying to automate SemVer rather than solving any of the other problems directly. I don't think changelogs should be automated anyway - I can `git log` that if I want a list. A changelog is an opportunity to communicate to a wider audience what is actually going on under the hood.
I quite dislike this style of writing titles. "Stop something". I seems very popular. It sounds very commanding and "I am definitely right about this". Why not write "In favour of something" or "A case against something" or something like that?
Why not be direct and advocate clearly for the position that you prefer? You don’t have agree with their position, but asking them to water down their words is weak sauce.
Oh please. Public discussion is always a balance, and to answer your question: if the content is nuanced, the title should be too.
If they mismatch some of your audience is unnecessarily put off.
Ill add: I am personally put off in the same way your parent comment is, because hard stances are usually wrong, and I like a bit of nuance in my life.
Not as bad as "considered harmful" imo but still mildly toxic. I think the point is taking one rando's personal preferences (I'd prefer we swap the order of A and B) and trying to make it sound like something more than it is.
> Not as bad as "considered harmful" imo but still mildly toxic. I think the point is taking one rando's personal preferences (I'd prefer we swap the order of A and B) and trying to make it sound like something more than it is.
Mildly toxic was the same but worse in my opinion.
I'd much rather people think deeply about summarizing their work. This helps others understand it but, more importantly, helps the developer understand what they did. If its hard to summarize, maybe it should be tightened up a little for instance. Enforcing a "schema" might help a tiny bit but also can cause people to check out a little as it can feel like just another meaningless process.
The thing conventional commits are really helpful for is continuous delivery. Every merge to main can be automatically tagged with semver and shipped because the thought that goes into tagging and versioning has already been done by the developers when they wrote the commit message.
I fully recognise that it doesn't make sense for huge projects like the Linux kernel to do this. But for 99% of projects conventional commits combined with semver vastly improves the release process status quo and makes it easy to automate.
I like relying on `git tags` even in continuous delivery situations. `git describe` is often good enough for continuous delivery versioning `v1.2.3-4-gabcdef` can describe a commit exactly enough to make git happy and is semver-like enough to set expectations, even/especially if new `git tags` are only ever inserted on human discretion (this is breaking change so I need to tag a new major now). The only real debate with `git describe` format version numbering is if to better conform to semver expectations the first dash should be a plus or not and you can do that change with a quick regex if you feel it is worth it for whatever is enforcing your semver expectations (ordering versions correctly in a package manager, for instance).
`git describe` is easy to automate for CD, but can leave version number decisions to people via `git tag` choices (and/or GitHub Releases) rather than trying to guess from commit history magic keywords.
I do this on my OSS projects to automate semver bumps and it's amazing! At work, I also enforce "tags" (not git tags, just strings in the PR title) based on who cares about the change and then generate changelogs for the respective teams based on those "tags".
Why does that need to go in the title though? Just put some magic in the commit bodies, if you want to do versioning in that weird way. Then you aren't restricted to a single word either.
“The audience of a changelog is entirely different than the audience for a commit log!
A changelog is user-facing”
I'd say that ship has probably sailed. Most companies are happy with “Bug Fixes & Performance Improvements”.
At least if they're not going to put the effort in, then a generated changelog is better than nothing.
The best thing that I'd used for auto-updated software with weekly updates was to prefix user-facing visible commits with "uv:" Then each week we search for them and either use the text as-is or massage it slightly. We even got it into the product itself in the Help/Release-notes menu.
Funny to ask to stop doing something I don't do or never even heard of. I typically only mark database schema migrations or other major things with special prefixes.
He confuses changelog with release notes. Doesnt know how to name commits and probably doesnt know how to name symbols either. Skill issue and he’s sad, now in public. Move on…
Odd. The main reason to use this style of commit message is for CI/CD automation.
EDIT: I didn't see this covered in the article on my first pass. It is covered though. My apologies.
The type of the commit informs the automated workflows how to handle the commit. This is why it comes first.
For example, if you're performing CD, if you only commit a bunch of `fix: ` then only your semantic versioning patch version number is incremented. If you commit a `feat: ` then it's a minor version is bump. `feat! ` is a major version bump.
Even if you're not using CD for releases, semantic commit messages are sometimes used to automate change log generation. Granted, your change logs should not typically include the Git commit messages themselves — those are developer facing, not user facing.
The article addresses both of these pretty clearly. Semantic versioning gets borked with reverts and the automatic changelog is targeting the wrong audience
The article is wrong about reverts (in my opinion). If a breaking change is introduced, and then removed, the removal should also most likely be considered a breaking change (both the addition and removal are changing your API). So it is correct that a major version bump should occur when reverting. Once a package has been published, the ship has sailed.
The issue is that if there was no release in between, or only a beta or similar, you now have two breaking changes indicated by the commits, although in sum there is none since the last official release.
That's true, but depends on your workflow and release strategy.
If you are releasing upon every push to main/master (following what semantic release and conventional commits provides you in terms of automation), then it makes sense to perform major version bumps for the reverts.
If you have a manual release strategy, then it might not make sense to use these tools in the way they have been designed.
If you have actual dependents in a SemVer fashion, then this isn’t useful for those still on the prior version. What you’d rather do is decrement the major version again because it’s compatible with the prior version again. Those dependents who already upgraded to the interim version have to consider another breaking change regardless.
And if you don’t have these kinds of dependents, then the versioning scheme isn’t important anyway.
release-please[0] allows you to do a manual version override in a commit, which would allow you to decrement the major version upon reverting a breaking change
I think that could be simplified, so the tool can tell that a commit is reverting a breaking change and thus the version should be decremented, but at least there's an escape hatch.
My apologies, I missed this on first read due to the indentation style. That said, I don't agree on the commentary.
Why on Earth are people not writing commit messages for their reverts? They should have semantic commit messages just the same as any other commit.
Unless the point is that they're not following per-commit CD, and if you commit then revert that commit before a release was made. That sounds like a process failure. Which of course, process isn't infallible, and neither is the automated version management. If you screw up, use an escape hatch — just like reverting a commit that had previously gone through code review and been merged.
Re: change log generation. The article says change logs shouldn't have commit messages. I agree. Many tools (e.g. Changesets https://github.com/changesets/changesets) use the semantic commit type to sort change log entries, but require you to write those user facing change log entries separately.
Use some convention for git trailers then. Having “fix” or “feat” in the commit title does not provide any useful information to someone scanning the log.
How... how is this not obviously the absolute very most useful information?
When I encounter a bug in a dependency of mine. Before I worry about submitting a PR, the very first thing I do is grab my version number and check the commit logs for fixes since my version number.
If I'm trying to decide whether I should bother upgrading, I scan the log for new features.
It's the title, not the details. The commit message body should contain MUCH more detail than the title.
If you don't like it because it looks ugly. Sure, that's subjective. And actually, I agree. Because it's standardized though, Git interfaces could even be configured to trim this off and provide different visual styles for the different kinds of commits. The types could be used as search filters too etc.
Now, I get people don't like the look of them. Neither did I when I first saw them. Then I started using them and found them useful.
It's fine, people have different preferences, it's just a convention and it's not going to work for every project. The article itself just doesn't seem to hold any water.
If one is writing trailers and custom formatters, then probably the information that the formatter uses should be even more structured that sticking it in the subject line.
If I'm at the point of contributing a PR to a dependency, I've already identified the root cause in detail. There's no way a change log should be going into that level of detail, or else you're just duplicating the Git log for no reason.
Will the change log make mention of fixing the bug? Perhaps. But I'm going to want to read the technical details of the fix to make sure they've specifically addressed my issue, and not just a similar problem. What is the performance impact of the fix? Are there security implications they've explained in the commit message.
I'm a software engineer, not an end user, I want the technical details of my dependencies.
Wait. i am already struggling with 50 columns for the title, now i have add this `<type>[scope]:` bloat to it? How are these people doing both "Conventional Commits" and 50-72 at the same time?
None of this is as good as "free array memory before it goes out of scope". This is better than `fix: memory leak` - which is what most people would do. It's also better than `fix: free array memory...` because `fix:` is redundant when you have a good message. I get people want to build automation around this stuff, but just do that in a footer of a commit message where humans don't need to see it.
Whole idea of CC is to write commits in away that is easy to generate change logs. If you utilize CC correctly or not, at best you get: commit log that harder to read, changelog that is hard to read and still requires you to write highlights (guess minor and path releases are fine).
Though the example in the actual specification, “fix: array parsing issue when multiple spaces were contained in string”, is more inconclusive (and frankly doesn’t really make sense as a description).
yep. I'm on the fence about types generally, but "fix:" saves/standardizes a bunch of phrases like "fix an issue where", "prevent" or having to invert the message by describing the solution instead.
You never need to write "fix", or "prevent" if you write a good message. Nobody says "fixed the car by changing the flat tire" or "Go prevent engine wear by getting an oil change today. You say "Change the flat tire", or "Change the oil".
You can do the same when you write commit messages. "Wrap user and account update in a transaction" - "Delete temp files after use".
I disagree. I think the commit description should be "stop the engine wear", not "change the oil". I much prefer commit messages describing why you did the changes, because the explanation of what you did is already there - the code diff itself.
I feel +/- line counts alone give a good sense of whether something is a fix, a refactor, new feature/component.
After that it’s great to know where is primarily intended to be impacted.
Then look for what the change is and does.
There are some good ideas in this proposal but the author is creating a false dichotomy by saying the current standard should be scrapped. Two things can coexist and both be good. I hope the author reconsiders their approach in their future promotion of their idea.
I totally agree with the sentiment behind this, so much so that I also made my own standard a while ago, https://commits-with-character.org/ an incredibly light addition to the Git Book guidelines, and also gives more priority to scope. It’s been working really well for me in my small projects.
Broken promise 1 (Automatically generating CHANGELOGs) is spot on, especially in a monorepo with multiple released artifacts. A single commit could be a breaking change in one package and refactor in another. But the changelog tools see one commit and include it in both changelogs.
Another issue is that once the commit lands on trunk, you can’t revise the entry without editing history. You have to remember to fix it after the changelog is generated.
Article is too opinionated IMO. I enforce CC on my projects because I don’t have the energy to police horrendous commit messages. It’s easy to enforce the CC format on the repo merge policy. I do it with the addition of a required issue ID as well.
If I only worked with seasoned devs, I wouldn’t use it, but that’s just the reality of my work. It also has a bonus of forcing AI agents to write in the same form as well instead of their random personal flavor. Precommit hooks stop everything before it gets in front of my eyes for review.
> I enforce CC on my projects because I don’t have the energy to police horrendous commit messages.
And does it actually accomplish that goal? I've been on several projects where someone pushed CC on the team with this reasoning. Every time my experience has been that you get the same crappy messages with a tag that may or may not be accurate.
BTW, AI absolutely knows how to bypass pre-commit hooks and will do so when they come up with some reasoning why their situation is an exception to the rule. I've watched them do it. The only way I've found to strictly enforce things on an agent (tests, linting, whatever) is to use a claude pre-command hook that will block git commit if the checks don't pass.
Keep a Changelog[0] maintainer here, weirdly seeing this while deep into work on way overdue 2.0 "release"[1], which should be out very soon.
I've had to contend with Conventional Commits both in the OSS world and at work as it proliferated from what seemed to me like robotic adoption by folks who were even loosely associated with the Angular ecosystem (remember that?).
I've always had a stance with KAC that folks trying to automate changelog creation (prior to LLM rise, mind you) were focusing on the wrong thing. I still think there's a fundamental difference in focus between what you write in a git commit and what you present in a changelog.
I know there are fundamental philosophical differences for folks who were used to HISTORY vs. NEWS vs. CHANGELOG but with the growing adoption of KAC-like CHANGELOG.md files and Release Notes (often not synonymous) I think we're thankfully past the weird era were maintainers dumped raw git log ranges between two tags and called that a changelog. I'm sure some still do it. But that's what Conventional Commits tries to replicate.
What's really odd to me is that this assumes (broadly) that every single commit in a repository is relevant to the eventual version release changelog (or release notes). Even if you assume some CC types get filtered and deprioritized from generated changelogs by some tools, it's still a huge miss on what communicating about a release typically means: these change likely matter to you as a package dependent or direct user, while others were omitted for good reason.
I'm trying to articulate that much more clearly in KAC 2.0 because there's a fundamental paradigm shift when a robot can now analyze recent work (yours or theirs) and craft changelog entries that appropriately shift the audience perspective from "git message for me/us in the future to understand this change" to "changelog entry for you/them to know what this group of changes means".
The issue with all of these schemes is less about the format and more about the semantics itself. What are all the actions that can be done to a codebase and what is a controlled vocabulary that encapsulates those? Then it doesn’t matter what system you use.
I spent some time recently coming to the conclusion that I did not prefer CC, but wanting some reliable structure. In the end, I found I was coming up with convoluted schemes that were getting in the way of actually solving my real problems and just settled on the tried and true:
“When applied this commit will...”
- Add <functionality>
- Update <existing>
- Refactor <while keeping same boundary behavior>
- Remove <some subsystem or functionality>
- Cleanup <documentation or style>
I don’t consider this to be a complete taxonomy, but it does let me get on with my day and covers most things, especially when combined with thoughtful commit messages.
There's no benefit to any of this. Just write like human. It will be clear if it's a fix, or a refactor, or ?. Typically it isn't just one of those things.
Making an opinionated absolute statement about something that's widely used and argued about isn't even logical. If nobody got any benefit, it wouldn't be a thing.
The CC standard may not be for everyone, but having some convention is often helpful. git log --pretty=oneline of structured format gives you broad filtering that's useful. Just writing like a human doesn't give you that ability and you're forced to read every line.
This is not without struggles. Many times the changelog updates are missed. You can try to catch this in code review, but that could also be missed. So you can try to automatically verify the changelog was updated, but you can't force that as a pass/fail check since not all changes require a user facing change. Or your project maintainers simply copy the commit message and paste it into the changelog, and at that point, why not just automate it with something like conventional commits?
Could/should the changelog be considered a first-class deliverable with care and attention provided? I think so, but I'm not in a position to exert direct control over that across dozens of repos and team members.
In my experience, LLMs are great at reviewing changelogs for potential gaps from a user POV (and even creating draft changelogs wholesale, if you're backfilling) based on git history.
The essay gives a bunch of reasons to drop the "type" altogether and just use "<optional scope>" as prefix. The type either doesn't really mean anything or is redundant when writing commit headlines as English sentences. In a message like "Prevent thing from happening" the verb "prevent" is already basically a synonym for "fix". Similarly "Add" or "Support" likely implies "feat"/"feature".
To some extent the "type" is simply about trying to limit/standardize the number of possible "verbs" to start a commit headline with, in which case Conventional Commits made the mistake of mixing verbs and nouns (fix and refactor are verbs but feature and chore are nouns) and adding distracting punctuation where English prefers none between the Verb and its direct object in a "Verb the thing" sentence. "Verb: the thing" only ever really looks awkward.
But also do we really want to limit the possible number of verbs that a headline sentence can start with when making commits? "Fix" and "Prevent" may often act like synonyms but there are connotative differences. In some cases "Prevent" may be a shorter way to explain why something needed to be fixed in a headline because "prevent" also says "stop a thing from happening that wasn't supposed to happen" whereas "fix" alone may not yield that extra context. The top line of a commit should be a short and sweet headline and sometimes the cleanest way to do that is to use the full gamut of English verbs at your disposal to tell the right story as quickly as possible.
In my experience both are more often in noun form with an article ("a fix" and "a refactor") because they are both sort of awkward alone. But sure I can appreciate that's probably how we got the set of words that we did that they were all picked as nouns and only in juxtaposition does it become more obvious to me that I think it might have been better to have picked just verbs. I can see the sorts of thoughts that led to "noun: verb the object" as a "good" headline format, especially when the "noun:" is considered the important part of the headline. But only before realizing it is generally redundant between "noun" and "verb" and "noun:" might not really be all that important (as the article points out flaws in).
I personally like it for one of my work projects. Its the only one we use it on. Its a repo with the android and iOS app. I like the conventional commits because when doing releases, I can look between two tags and pick out what I need to put in the "What's new in this release" section. I try to not just do the normal, "bug fixes and various improvements," in that section, but what we actually did. Also helping make clear what went into a release branch/what needs to be cherry picked into a release branch. I also don't automate the generation of the "Whats new" section. I just take a look at all the commits between tags.
I stick with Conventional Commits as a standardised way for communicating changes. Most contributions in the form of open PRs are squashed before merging into base branches. A team member can commit as personally wished in the feature branch. I still use CC for my own sanity and I’m very glad I do comparatively to what usually find in others feature branch commit histories.
Without standards most developers I’ve seen are very careless.
I generally work with changesets to curate changes at the time of contribution in accordance to semantic versioning.
This is the kind of thing where the extent to which people get worked up over it far exceeds the benefit of actually doing so.
Is it theoretically possible that people can write good commit messages without a framework? Yes. Does that work out in practice across entire teams? No. Any kind of a convention for commit messages is an improvement.
Beyond that, type > scope vs scope > type or whatever is splitting hairs. The juice of that debate is not worth the squeeze.
I have long despised Conventional Commits for pretty much these reasons. Yes, it’s structure, but it’s not useful structure. Of the five things it claims to enable, three are nonsense and the other two are actively bad.
And it’s ugly.
(But I suppose I am talking primarily about the first line part. The “BREAKING CHANGE” bit is potentially actually useful, though being incompatible with git-interpret-trailers despite leaning on git-interpret-trailers for other footers seems a bit crazy.)
I like using conventional commits but I’ve often wondered if some sort of tagging/labeling using git-notes wouldn’t be better: https://git-scm.com/docs/git-notes
I’m just unsure that the short title is the ideal place to put this kind of tagging info: the kind of fix, and optionally, the relevant component(s). I find sometimes that can take up the majority of the title.
A forge could consume the git-notes and decorate a commit/pr accordingly. Heck, GitHub PRs already have a labeling system in place, just have to add some glue.
I agree with the sentiments -- knowing the scope of a change is more essential, in most cases, than knowing whether the change was a "fix" or a "feature" -- but much of the scope in a project can be gleaned from the files which were touched. So the use case of, "i need to quickly at a glance determine what commits touched what parts of the project" can be answered with some variation on
I used and use Conventional Commits in private and professional projects and I think they have a place.
That being said, when I finally committed to using them something in me broke. Most of my career work was double faced. One face to the customer where you need to keep Tatemae under all circumstances (UI, Tickets, etc.). The other was the code and commit messages where you still kept your professionalism but you could be open and speak the language of your peers.
This time is over, maybe for good, but a tiny part of me misses it.
I see more of these conventional commit-style comments recently and it feels like coming from Claude Code etc. It's a bit unsettling that not only training data but also random lines in the default system prompt affects this kind of software development norms in subtle and pervasive ways.
Conventional commits (especially with git emojis) show at a glance the blast radius of the change (eg whether it breaks the product itself or just some internal dev tools). Emojis help immensely when looking at dozens of commits at a time.
That information should still be in the commit messages. "No functional change intended." appears widely in FreeBSD commit logs when code is being refactored (or, rarely, restyled).
And the issue isn't whether you can remember what you changed yesterday; this is largely about making sure other developers can quickly identify relevant commits. If you're a solo non-OSS developer, this is entirely relevant to you.
From my own experience in big tech and various OSS things, the crux of this piece rests on this whole scope not being focal point misses that in general the other practice is making sure PRs have limited scope in general so it's marginalized, i.e. reducing blast radius or too big of a change at once by design is implied.
As a reviewer, I love type first because it sets expectations for what I'm going to look at. Similar if I'm bisecting or doing other history operations.
I don't do automated changelogs or versioning but it also makes it faster for me to do so.
I really dislike focusing on issue ids. I only want to jump to another tool if I need more information, so put it in the footer for after I've read what is there, like a front page news article giving you the option to go to the back to read more. Worst case that I've seen is people that think the Issue ID is all you need.
I've been a developer for 40 years. Used VC since rcs. If only people had such pedantry for the actual code, the world would be a better place. I look at the author way more than I read a commit message, next the date and then the code. Never much else.
I think some structure in commit messages is helpful, but not to the point where it gets in the way of effectively reflecting what the commit contains, why it was done, and any comments for future reference, e.g. potential regressions.
the article mentions three types of stakeholders: contributors, debuggers and incident responders. it entirely fails to mentions consumers, who mostly care about backwards compatibility of changes, and thus about the type of a change. once the type is established, e.g. a change is breakimg, the consumer next cares about the scope to make downstream adjustments.
the part about broken promises regarding breaking changes is _kind of_ fair, but only assuming tooling isn't able to track reverts. accidental breakages occure with every approach and better to have an approximation than no information at all.
I'm not getting why "scope" gets any priority at all. Are filenames not attached to a commit for those exceedingly rare cases where you're looking for which files are thrashing?
Thank you for the thorough write up! I just thought about this today and needed the set of arguments and examples you brought forward. Especially the pointers to NixOS et.al. were informative. Thanks!
Definitely agree that generating change logs from commits leads to confusing change logs for people that expect to see what changed between versions. A big long list of commits is too granular. A curated and summarized list of changes is much more in-line with what most people expect when reading a change log.
I have never once considered conventional commits to be about human understanding, but more about automation (release notes, changelogs, and workflows).
The commit description and the pull request are for humans.
Commits should have no information in them. Teams should be aligned on the design of their software, and all the information about that software should be apparent from its source code.
Why not both? Scope is definitely important. But type is also important:
- prioritization: bugfix > docs
- visibility: are we spending more time on bug fixes or features?
Also, we let AI write the code, are we STILL writing commit messages by hand??
Mine is “ticket id - Imperative phrase”. Then I write a “why” description of the changes if needs be. As for personal project, I quite like the scoped commits style.
I don't care much what it says as in "fix", "chore" etc, but for me the main benefit is breaking changes indicated with "<type>!", something like "feat!: ...
".
This makes neovim plugin manager highlight the change differently which brings attention to it when you update stuff.
So please do use it instead of complaining!
I do like the suggestion of
scope!: ...
if it will be treated the same way with breaking changes reactions.
Asides from the well made points here ('scope is more important than type' etc).
> something like fix, feat, chore, docs, or refactor
'Docs' are also part of the program, they need fixes too, and features need docs. If the docs don't match the features because they're not being updated when the code is, the docs are a lie and waste other developers time.
Also if you were writing a standard: why would you randomly abbreviate 'feature' but not 'refactor'? That sounds like a nitpick but standards require great thought, this is a bit of a smell that there hasn't been much thought into designing 'conventional commits'.
Finally: the name 'Conventional commits' is a land grab (reminds me of when someone made a JS Standard and called it 'StandardJS', ignoring every existing popular standard). From the article, the *actual* convention is 'scope: work"
terrible suggestion, people are awful at writing commit messages and the type is really helpful when you're reviewing history and want to know things at a glance
You know what lacks focus? This whole article. Oh, and also individual styles of commit messages. You know what’s a waste of everyone’s time when you’re building technology in 2026? How to write and summarize what you’re work is doing.
Conventional commits doesn’t make a promise, it’s a specification. Words have power and meaning, this viewpoint is about more than a spec, it’s about the popularity and relevance that they clearly despise.
Conventional commits are like BEM syntax for CSS, it works if you pay attention. It’s structure around what developers do, which thoughtful limits. What is the problem with that? Is it the only way? No. But to say it’s encouraging the “wrong things” pshhhh
This author’s take is dog water for coders who yearn to be controversial and (clearly) focus on the wrong things.
I really dont care about commit messages. Just create strict rules for branches that contains issue nr + description, and squash all commits on merging the PR.
I'm not saying that conventional commits are God's given best way to structure a commit message, but they are a defined structure, and I find it much more effective and important that some expectations be set around commit messages, and I think conventional commits are as good as anything.
Like the author is making a big deal that they think scope is more important than type. I may tend to agree, but I think the difference between "fix(compiler)" and "compiler fix" is not exactly a hill I'd be willing to die on.
The tech industry has tons of things that became standards even if they weren't optimal. E.g. if one were starting from scratch I think any sane person would argue JSON should support comments (sorry but Douglas Crawford's rationale for not including comments never made sense to me), better defined numeric formats, etc. But it was better in many contexts than what came before it, so it became the standard. I could believe that there is some other format that differs a bit from conventional commits that is a little better, but not really better enough to want a whole other competing way of structuring comments.
reply