News for hackers. Stuff that matters.
Re:Why did Heroku fail? by craigkerstiens on Friday May 13, 2022 @07:15PM (#31373524)
This is a bad take without much internal context. Heroku is not thriving as it stands today. It takes a lot just to run Heroku and the engineers still around care deeply around it. I'm excited for a lot of the new offerings like Fly and Railway, but they aren't operating at the same scale as Heroku.

It does feel like there continues to be a fading at a leadership level about developer experience and what it means. There may not be strategy for what makes Heroku the next Heroku whether engineering or product, but the team that is still there is a solid set of engineers, not 100% there for Salesforce internal stuff, and kudos to them for still keeping afloat what is one of the better ways to deploy an app.

Re:Why did Heroku fail? by hthrowaway5 on Friday May 13, 2022 @08:30PM (#31373955)
(Former Herokai) Heroku is beyond the point of recovery at this point (barring a sale to Github or something). Those dedicated engineers need to give up because at this point they're harming everyone:

* The industry – suffering because Heroku holds market share that should go to competitors capable of innovating.

* Customers – suffering because their platform will die but their vendor isn't being transparent about that fact.

* Herokai – suffering because they could be doing something better with their careers, even if it is on another sfdc product.

Make sunset plans now and let customers know what the timeline will look like. "Trust is the #1 value" at sfdc, right?

Re:Why did Heroku fail? by dubswithus on Friday May 13, 2022 @07:40PM (#31373686)
For Fly.io I need to create a docker container? Why would I want to do that when I can just push to Heroku?
Re:Why did Heroku fail? by jolux on Friday May 13, 2022 @07:57PM (#31373781)
I can see this argument but at the same time Dockerfiles for my preferred framework (Phoenix) are more or less muscle memory for me at this point. The better question is: why don't frameworks come with a Dockerfile by default? The effort:reward ratio on containerization is huge, it gives you your app in a runnable, dependency-free format that you can build reliably and deploy on an extremely diverse set of platforms.
Re:Why did Heroku fail? by jakear on Friday May 13, 2022 @11:55PM (#31374952)
Dependency free? Isn’t Docker a dependency itself? And it’s the most annoying one — no other dependency constantly keeps a core going at 15% on my machine, or considers not upgrading a premium feature, or incessantly interrupts my work with NPS pop ups even when I’m doing nothing related to Docker.

If anyone at Docker is reading this, I mark “100% no” when you interrupt me with pop up asking if I’d recommend your product to a friend when I’m in the middle of working on something entirely unrelated to your product. I don’t know anyone who wouldn’t.

Re:Why did Heroku fail? by jolux on Saturday May 14, 2022 @12:38AM (#31375129)
Docker Desktop is a development dependency, but you can also use Podman or minikube. I just use Podman these days, I prefer it. The container format is all standard these days, you don’t actually need anything Docker to use it. Fargate and Fly don’t even actually run the container directly, they convert it to a VM with Firecracker first.
Re:Why did Heroku fail? by dubswithus on Saturday May 14, 2022 @12:26PM (#31379555)
Running containers make me feel like I need production hardware on my development machine. It's slow, clunky, and uses too much memory. I'd rather boot it up without docker.
Re:Why did Heroku fail? by jolux on Saturday May 14, 2022 @06:08PM (#31383110)
There's no requirement to use Docker while developing, it's just useful when you need to deploy.
Re:Why did Heroku fail? by bastawhiz on Saturday May 14, 2022 @01:17PM (#31380127)
> but they aren't operating at the same scale as Heroku.

As someone who said "I'm sure it can't get any worse" with heroku for many years and was proven wrong over and over, I genuinely don't know what you mean by this. Lots of stuff just doesn't work. Updated buildpacks don't get released on a useful timescale ("just make your own" isn't an answer). The platform is plagued by bizarre performance issues and noisy neighbors, and the pricing doesn't support it. In my time on Heroku, I couldn't break three nines of uptime one year entirely because of platform issues, including one incident where the DNS provider Heroku uses just stopped working in Israel (which Heroku gaslit me about for over a week). I don't know what "operating at scale" means to you, but if the service is consistently bad compared to the competition, it doesn't matter what scale they operate at. It's a meaningless metric.