[syntheticnature]

I expect it was thought through, and the thought was "a few hundred thousand people use this service still, it's running under one employee's desk* and is completely unfunded, no effort above the minimum is to be expended."

*Not sure if this was explicitly said or someone's joking claim during a past discussion of the shutdown.

[aimthrowaway5]

The final user numbers were way less than you would guess. It was a service built to handle tens of millions of users serving thousands. Scaling down that much is as labour intensive as scaling up. It's not just a launching all the services on a handful of VMs. The time had passed to make it a WhatsApp etc competitor.

[ancarda]

>Scaling down that much is as labour intensive as scaling up.

Could you elaborate on this? I'm very interested to understand the challenges of downsizing a cloud service as I've never heard of that happening, and am surprised to hear it's difficult.

[lev99]

1. Everyone that wrote the code has probably left the company, or is working on more important projects.

2. The code was written in 90s and 00s.

3. I wouldn't be surprised if AOL owns the hardware AIM is running on. Deploying on smaller servers would mean a complete system replacement, which always requires large amounts of regression testing.

4. Big services usually have more dependencies than small services. Each dependency adds complexity and costs.

It probably is as hard to down scale an old service as it is to upscale a new service. It's probably less hard to downscale a new service than upscale a new service.

Also, if your users are dwindling (and not paying) it's less important to give that software attention.

[syntheticnature]

Down to just thousands, not even tens of thousands? Wow. I knew about three others still using it (and I had it around for them), though only two of them messaged me with any regularity.