Bitbucket Server is written in Java and bitbucket.org is written in Python.

I was dumbfounded when I found out they were two separate codebases.

(hate it when companies do this)

https://www.mercurial-scm.org/repo/hg/

Mercurial is very important for Facebook, Mozilla, Google, and others. The disparaging "legacy" monicker might be a good way to feel justified to not have to learn about it, but it's not dying software that is no longer getting updates.

Why would you want to use such a limited diff tool at all? Personally, I use the best diff tool that I can find and so far that's BeyondCompare from Scooter Software.

I wonder if they were slow on updating their site because of Sourcetree, which I personally love (especially since it is a cross-platoform [Mac/Win] GUI client).

A large enterprise that's fully exploiting Jira will set it up for non-developers, indeed they'll set up Jira primarily for their non-developers. HR will have an HR project with the following issue types: New Hires, Employee Disputes, etc., each with workflows with statuses like resume received, contacted to set up a first interview, first interview set up, offer extended... etc. And then you can use all of Jira's existing reporting schemes to help management figure out, if I need to hire someone new for some type of position, how long does that take? Are there differences between different positions? How long does each stage in the process take? Why? Does the aggregate data show some kind of bias - gender or otherwise? How do I make this faster? And so on. The kinds of questions that you can ask when you're an enterprise with established process and not a start-up with the entire company fit into one room.

Internally, even within our engineering division, we've been using issue linking to other projects to get better visibility into why various work items get stuck. If I'm dependent on some internal service for my work item, and that service becomes unavailable for some period of time, I can grab the Jira support issue tracking the unavailability, mark my work item as blocked by that issue, and then that's visible all the way up the enterprise hierarchy. Similarly, I don't have to ask the people responsible whether they've fixed their system yet, my own item updates automatically to show that my item isn't blocked anymore, and I can go back to work.

The main difference between Jira and Gitlab is that Jira is a workflow management tool and Gitlab is a software management tool. Gitlab looks fantastic if I want to start a new software company and have everything be right there in one window for me. But if you put all the software parts in front of all your non-software teams, they'll balk. It's just not relevant to them.

Microsoft has a tool much like Gitlab called TFS, which has been around for far longer than Gitlab has. TFS is also designed to have this one-stop-shop UI, and TFS also never became an enterprise workflow management tool, even though it too has issues and workflows and customizations up the wazoo. It's a product that's targeted to software developers and it knows it, and that's why hardly any large enterprises actually use it unless they have old .NET projects that have been on TFS forever.

This is the benefit of Atlassian's multiple-product model. Jira is management-first. Bitbucket is software-first. Confluence is customer-first. In an all-in-one UI, you have to pick who's first.

The real problem Gitlab faces is how to retain software projects in growing companies once those companies expand and become large enterprises, and need to start to track more generalized workflows. Because if you had to ask what's most important in the Gitlab UI, workflow and reporting is definitely not it.