Hacker News Reader

Comment by smaudet | Hacker News Reader

Hmm - review is not a part of git, in fact it is antithetical to git (concept from cvs/svn with centralized systems).

A pull request, literally, need only be moving code from one repo to another.

The signaling mechanism etc. are not and shouldn't be specified by git. How code moves around should be dictated by your organism, copyright, access levels, not some guy's interpretation of what "pull request" means.

That means, yes in some instances there is no 'master' version of the software, there are multiple flavors with different release characteristics.

E-mail I suppose is an adequate communication medium, if people want a specific review mechanism and triggers adding one to source hut (assuming FOSS) shouldn't be hard, perhaps the authors intention is to make it difficult to regress to a centralized model.

View on HN

>Hmm - review is not a part of git

Email-based review tools are built-in to git. git format-patch, send-email, am, imap-send, etc.

Those are tools to facilitate moving the code from one repo to another, not to facilitate a review of that code. So I think the GP's point stands.

I mean, it facilitates code review as well. You just reply to the email. This is how Linux is developed, for instance.

Yes, but you do not reply via git. You review the patches in the MUA of your choice (outside of git) or in editor or IDE of your choice, after applying them in your local repository clone (applied maybe by git, maybe by the "patch" command, but viewed and evaluated completely without git), then you compose a reply in your MUA or maybe have a direct chat with someone about it (again, outside of git).

Git only comes in play when the review is finished, and it's time to commit the changes to your repository. So I claim that git does not have tools for reviewing pull requests or patch sets, only for managing them on either end of the pipeline - creating them, sending them, and applying them.

Which, to be clear, is not a criticism of git, as I do not think git should grow such review tools.

Alright, sure, but you're picking nits and I'm not sure what your underlying point is.

I do not believe it's nitpicking. I was supporting the post you replied to (https://news.ycombinator.com/item?id=33313036), which claimed that git indeed does not have review tools, and that the review tools/flows are and should be external to git. You claimed otherwise, and I argued that you were incorrect in that claim.

Isn't that what discussion board is for? I agree that this particular topic is not life-and-death important, but still worth discussing, IMHO. If you think otherwise, just walk away, you do not owe anybody anything.

It isn't picking nits, git does patch management, there is nothing and shouldn't be for review.

It matters because where you assign your responsibilities is the difference between spaghetti code/architecture and a well designed one. If you think this is nitpicking, I encourage you to (re)view the general sofware concepts of cohesion and coupling.

(Part of) gits success is it is a tool focused on doing one thing well - I believe Linus is on record as saying the other tooling (commercial/svn) largely failed because they were not really focused on source control i.e. patch management. Its part of the reason he constructed it in the first place, the tool he was using was overcomplex and bad at doing simple things.

Github's success until this point has been largely that it focused on building review infrastructure. It's based on git, giving it flexibility akin to being based on e.g. http, but it's core is review and facilities in support of that e.g. code search. It's a very particular type of review aimed at a specific audience, but it's not git, nor should it be.

Reply on news.ycombinator.com