Why was this apparently easy enough to do that you could do it by accident?
Why was it impossible to recover from what was clearly an inevitability?
I feel bad for whoever fatfingered the "delete literally fucking everything" switch but I'm not even sure an individual can actually be said to have been solely at fault for such a colossal and systemic screwup.
The converse of this is “lock everything down so that de facto nobody can make any changes to it, especially those important new features teed up for the next sprint"
I don't know how many employees KPMG has but many of these chats were probably destined to be deleted anyway because big companies typically have short retention windows to limit litigation discovery. The main exception being material on litigation hold, which makes me cynically wonder what KPMG "accidentally lost".
My limited experience in MS Teams has found a number of features that have no version history/logs or even basic undelete/archive functionality. It's unreliable for collaborative work.
Teams Planner - what happens when a Team member deletes tasks or renames tasks? Good luck finding what happened to the task or who destroyed your project plan.
Wiki has no clear way to track changes. If someone deletes wiki or pages, chances are they are gone for good.
Wiki and Planner have plenty of horror stories on the MS Uservoice. Some requests to fix this basic functionality are over 4 years old (still being ignored by MS).
People said "store your data in the magically overpriced cloud, it's much safer than running your own servers! And also automate everything!" However by now I am convinced that people are more likely to lose data clicking the wrong thing in the horrifyingly complex mess that is AWS and the like than a single dedicated server becoming unrecoverable.
Maybe instead of this "automate everything" we should pick and choose which things should be automated, and what should be performed manually. Probably shouldn't have automated data loss.
Other professions have known the risks of a dangerous thing being too easy or too fast for ages.
And I don't mean "slap another password prompt and a second confirmation dialog on it". I mean: want to perform some destructive/dangerous action? You'll have to perform it for each server/customer individually. Yes it's going to take days, but at least you'll only fuck up for some of them before you realize your mistake.
"Really? But can we please automate that?!" Sigh. Just make sure it stays slow and still takes days, so you can stop it before everything is gone. Incremental rollouts are the second best thing.
But it is pretty ironic accountant firms sell technology consulting to their customers while they haven't automated their own core business. Audits start at 100k for SME's because it is still billed by the hour. On top of this even the CIO doesn't understand IAM.
7 comments
[ 447 ms ] story [ 1521 ms ] threadWhy was this even possible?
Why was this apparently easy enough to do that you could do it by accident?
Why was it impossible to recover from what was clearly an inevitability?
I feel bad for whoever fatfingered the "delete literally fucking everything" switch but I'm not even sure an individual can actually be said to have been solely at fault for such a colossal and systemic screwup.
Teams Planner - what happens when a Team member deletes tasks or renames tasks? Good luck finding what happened to the task or who destroyed your project plan.
Wiki has no clear way to track changes. If someone deletes wiki or pages, chances are they are gone for good.
Wiki and Planner have plenty of horror stories on the MS Uservoice. Some requests to fix this basic functionality are over 4 years old (still being ignored by MS).
This is the flipside of automation.
People said "store your data in the magically overpriced cloud, it's much safer than running your own servers! And also automate everything!" However by now I am convinced that people are more likely to lose data clicking the wrong thing in the horrifyingly complex mess that is AWS and the like than a single dedicated server becoming unrecoverable.
Maybe instead of this "automate everything" we should pick and choose which things should be automated, and what should be performed manually. Probably shouldn't have automated data loss.
Other professions have known the risks of a dangerous thing being too easy or too fast for ages.
And I don't mean "slap another password prompt and a second confirmation dialog on it". I mean: want to perform some destructive/dangerous action? You'll have to perform it for each server/customer individually. Yes it's going to take days, but at least you'll only fuck up for some of them before you realize your mistake.
"Really? But can we please automate that?!" Sigh. Just make sure it stays slow and still takes days, so you can stop it before everything is gone. Incremental rollouts are the second best thing.
[1]: https://news.ycombinator.com/item?id=24196131 Canon
[2]: https://news.ycombinator.com/item?id=24229864 Adobe
https://microsoftteams.uservoice.com/forums/555103/suggestio...