Github Redesign (2019)
43 points by decentstates
43 points by decentstates
I suggest the conversation steer away from the controversial last point where he flattens the UI quite a bit to make it less "dated", as the rest was quite good up until that point, and that deserves being discussed.
I was with OP for around the 7th change. Near everything after that feels like an abject disaster. I do like the idea of having a sneak-peek at repo activity, but blanking last changed and last commit messages feels like throwing out a lot of functionality for no reason.
IMO Github's UI design is pretty evergreen. Its strength is that it doesn't need to change. The problem is more, that with all the JS magic they did over the years it's slow as hell and tends to fall apart if something goes wrong.
Agreed, up through the seventh item I was thinking "wow, this is so much better", but from 8 onwards it feels pretty incompatible with how I usually interact with the github web UI.
Same, losing the commit info next to the files doesn't feel right. Contributors, activity and language stats don't change often and aren't that useful after the first visit, but commit messages next to files/folders are a good visual indicator of what's changed recently.
I also wasn't a fan of applying the flat UI trend - having gone through the post explaining information hierarchy and flow, the author then threw it all away at the end. I found the flat version a lot harder to read since you need to work out where the headers and page divisions are rather than having a visual indication.
I’m not sold on flattening the Code subtabs entirely; perhaps the problem is rather that it was poorly executed. “Code” itself is a tree/blobs view, then commits, branches and contributors are clearly Git things, license was definitely in the wrong place (it’s repository information, like the language stripes below), and Releases was badly misplaced (which people have long complained about) since it’s not a Git thing but a GitHub thing. Releases should definitely have been a top-level tab, but perhaps Code could be a reasonable tab with Files/Commits/Branches/Contributors subtabs that are actually subtabs rather than the mess that they are.
Problem 8 is very, very wrong. “Without really losing anything” is nonsense.
Problem 9 reinstates the language statistics from the repository home page that were removed without remark in Problem 6, and the commit information removed in Problem 8. It is, to some extent, fixing problems that it only just inflicted (though having a page with recent commits and statistics and releases—absent from his design—could be useful). But there’s a big problem with such an invasive change: it practically breaks a lot of existing links. People rely on README files being rendered at the repository root URL. You must have that at least present on the repository root, whatever else is there. Not necessarily a blocker.
Problem 11… well, the clone link and instructions being hidden is a big deal. What’s there at present will be an utter, dismal failure in any usability testing. Something should have been done about that long, long ago. In the traditional “here’s a printout of a screenshot; you want to achieve X; where might you click to achieve that” form, I would expect a success rate of under 10%, quite possibly 0⁄100. Seriously, if you don’t know the answer because of having tried to find it at some point (probably because you remember it used to exist ten years ago): pick one and try to find it now. I expect you to fail. The feature might as well not exist for how well they’ve hidden it.
Dated look: frankly, just go back to the 2013 screenshot and see how much better it was. The subtab problem wasn’t a problem, because the top-level stuff was a sidebar, and the subtabs that were there were in a semantically-meaningful hierarchy: Releases didn’t exist yet, and they weren’t shoehorning licence information into the wrong place. Most of the problems identified started in 2015.
For the last commit touching files (problem 8), how do you use it? I feel like I spend a lot of time having my eyes drawn to it and not benefitting, because top level files often don’t change much.
I think most things I could get out of that information I’d be better off reading the linear commit history, or cloning the repo and running git log or git blame on particular files.
Problem 8 is very, very wrong. “Without really losing anything” is nonsense.
I don't think you really lose anything, at least relative to how much of the screen the information occupies. If you really want to keep the information, it could be collapsed into the short "last modified" label; just make it link to the latest commit it corresponds to and put the shortlog in the tooltip.
More useful than "the last commit" on any individual file or dir would be aggregate "recent activity" as a histogram or something.
On the front page of a repository specifically, yes, it’s arguably a proxy for that kind of recent activity, and you could present that information more effectively. But “last modified” alone is often too coarse, and splitting it up into when each file and directory was last modified can be useful. It depends on what the activity is.
Once you’re actively browsing the code, many people definitely want it frequently.
I think ultimately it still seems like an odd thing to constantly want to reference. You only get the last bit of activity, which may hide a dozen other extremely recent commits you would be interested in.
You also get weird artifacts from this, like a well-meaning commit that touches most of the repo permanently tagging quiet parts with something like "fixed line endings" until someone has to work on those parts again. I guess I don't know whether GitHub respects whatever that git file for masking out such banal commits is called, I guess that's the "solution".
I agree that the economy of space argument for dropping it applies mostly to the "landing page" for the repo
Just to make sure that I understand your reasoning, mainly because I am interested in understanding how UI/UX affects people. If once you clicked on any of the files to start "browsing" through the files, it would open up a new view basically mimicing the current file browser view on GitHub, then would that make it more usable to you again? Because I assume to browse files you are generally going to have to click into the file tree at least once anyway, so not having this info present at the highest level would not be damaging, right?
This was discussed long ago but I feel it is time for a revisit - GitHub's UI has become so aggravating.
I like the penultimate version, before removing borders etc. I thought I hated the version without last commit messages, but it's actually nice (to my eyes) as long as:
But yeah, the final version undid much of the "progress" haha.