15 comments

[ 3.9 ms ] story [ 48.7 ms ] thread
I refuse to believe that this was at all necessary with all of the enterprise chat applications out there. Uber is big but is in no way an anomalous case.
Yeah, this excerpt is hilarious:

"...our team transitioned the company to a new solution capable of reliably delivering over one million messages per day to tens of thousands of users, all in one unified chat environment."

Yep, totally no other chat apps capable of those incredibly high numbers /eyeroll

(comment deleted)
enterprise chat products are expensive and don't fit at uber scale
Unless data privacy is an issue; if you started your company with questionable means, keeping your data and internal comms completely private could be a (paranoid) priority.
So, an on premise enterprise chat? Something like HipChat maybe?

Less sarcastically, I don't actually think Mattermost was a bad choice. Sure, it sounds like they overengineered the crap out of it, the it sounds like it was fundamentally solid.

(comment deleted)
(comment deleted)
TL;DR: They used Mattermost [1] to chat/share-stuff and Puppet [2] for deployment.

  [1]: https://about.mattermost.com/features/
  [2]: https://puppet.com/
Screenshots look just like Slack or Discord. Why didn't they use Slack or Discord? What does Puppet have to do with any of this?
For one thing Slack takes and exceedingly long time from app open to being able to see messages. The guys over at Slack spend more time accommodating for loading catch phrases like 'you look nice today' than actually loading faster..
Because Slack is really expensive. Perhaps they concluded that an in house solution would be cheaper for them long term.
Likely privacy, cost and performance.
Slack is for "small" teams. They've got 10,000+ people in a 'group' since this isn't just for engineering staff, but also for drivers, customer support reps, etc.

These guys aren't split into lots of multiple teams under one big company, they're all just one 'team' with people leaving and joining fairly often due to driver turn-over.

Slack would cost ~$100,000 a month at those numbers as well (given enterprise level discounts).

Unexpected lazy solution.