Atlassian has discontinued their on-premises chat system in favor of Slack.

This is not a bad thing, overall. Slack has a better user experience and is much more popular. As part of this they have ensured that all of their tools integrate well with Slack. Unfortunately for me, I manage several installations of HipChat as part of my job, and they exist on airgapped networks (or at least there is a constraint that no project-related data traverses the internet). So this is a bit of a problem. I foresee that I will be spending a reasonable amount of time migrating data that's locked up in HipChat and integrating the other Atlassian tools (JIRA and Crucible mostly) with whatever we choose as our new chat system.

There is hope though. Mattermost's CEO made clear on HN that they are interested in making it easy for those with on-premises requirements to make the move to Mattermost. This raises the question: why didn't we just use Mattermost from the beginning? It's an open source tool, and mostly entirely free to use (they lock some of the features specific to enterprises, e.g. LDAP authentication, behind a license fee).

The other thing about open source tooling is that no one can take it away from you. Once you are running a particular piece of OSS, you will be able to use it in that form in perpetuity. You have the code, you have a license to use it (GPL, MIT, BSD, etc.). It can't be pulled out from under you in favor of a cloud service regardless of how popular and high-quality it is. I have expressed this advantage before - mostly as a reason I'd prefer not to learn a proprietary programming language or configuration management system. I don't want to have technical knowledge that's rendered useless by someone else's business decision. The bottom line here is that it's good to have a specific example to point to. I can't move to the cloud, and because my license for HipChat will expire, I can't just continue to use it. Atlassian has created some unnecessary work for me with this decision - work that I would not have to do, money that my company would not have to spend, if we had picked an open source setup from the beginning.