Instead of porting to Python 3, I'm porting away from Google.

Let’s say you had an old large code base in 100,000s of lines of code that just works and has been mostly untouched for years in eol/maintenance. you expect there to be budget or time to rewrite it into python 3 for literally no reason other then “2.7 is deprecated”?

It’s far cheaper, potentially millions cheaper, to just change clouds/providers and find a way to continue on 2.7.

I've migrated 200k LoC python 2.7 codebases to python 3. It certainly wasn't painless but tools like black, flake8, 2to3 and other stuff helped. It's not impossible and can be executed in 2-4 weeks

> you expect there to be budget or time to rewrite it into python 3 for literally no reason other then “2.7 is deprecated”

2.7 is not just deprecated. It's dead. There's no more updates. It's not "millions cheaper" to not migrate. I posit you'd save a LOT of money from the sheer performance improvements and bug fixes from migrating than anything else. In my own case we saved a few hundred dollars a month from not having to spawn new containers and EC2 instances

This is what happens, when no one is maintaining software.

Security is more important that anything, certainly than having time or not to update your code.

You're 100% in the wrong here, and Google is 100% in the right, and I hate Google more than spammers, more than pretty much any other company there is.

They're not wrong here, and it hurts me to support their stance.

Fix your stuff. Stop blaming others. I doesn't matter why you're in the situation you're in, you're wrong to be there. No matter what.

If you think security is an afterthought, you're doing it wrong.

this also really highlights the problem with python's 2 vs 3 split: there is a large body of work that will never be updated and will forever run 2 including actual web/publicly facing code. And saying "just upgrade because security" isn't going get much of that code upgraded.

This sort of thinking, is why places get hacked. You don't know what the access vector is. Stop making excuses. Secure your shit.

I would be so happy if a law was passed, and such actions were criminal.

Your code took a month, not everyone’s will.

so again something is just running and working why spend the money and or time when there are potentially far more important things to spend limited resources on?

op even said “yes I’m gonna but I don’t have the time” be

Like you mentioned, the docs were really shoddy when I needed them. Certain conventions felt half baked compared to AWS, and I don’t even like AWS to begin with. Overall it seemed like the integration of services was much worse than AWS, though. Administration of different services is like visiting separate worlds at times, like the teams might not be on the same page regarding design and technical language. The other possibility is I’m just a little dumb and don’t pick up on the patterns — I’m pretty open to that.

It’s a shame, I wish there was a cloud provider that I actually liked and at one point, that was GCP.

I'm curious what it is about Azure that puts you off.

I’m perfectly happy with AKS. Their Postgres product is great (formerly citus). But at the the end of the day, they still want to sell Windows, Exchange, SQL Server, Active Directory, etc. That’s a choice I will probably never make, so it just feels like a mismatch.

Granted my experience was inside Azure so we had the benefit of vertical integration.

I've spent very little time in the Microsoft ecosystem, but several years ago I was on a contract with a Microsoft shop and I found the MS server-related offerings to be pretty solid (granted, I was not the one paying for them).

The one thing that still shocks me (and maybe it's changed since I last checked, but I'm guessing not) is their pricing model around some offerings that are applied per processor core. Holy moly! Need to blow a lot of cash really fast? Then MS SQL server is for you. Fortunately (/s), it was a government contract, and at that point in the fiscal cycle they were desperate to blow cash so they didn't get less of it the next cycle.

That's always seemed like a perverse incentive to me: the people who do more with less resources get fewer resources the next round. Wouldn't it make more sense to give them more resources and incrementally expand the scope of their duties? They're clearly doing something right. Of course, it's always a balancing game. Perhaps the reason they were so efficient is because they had a limited amount of work to be done, and with more work they may not have performed with the same efficiency.

Perhaps the reasoning behind such a policy is that they're not generally looking for super heroes who might burn themselves out after doing amazing work. They're looking for people who are okay being a little bored, but who will reliably get a specific job done.

Comparable infrastructure on Azure somehow ended up costing that team around 60% of what their AWS setup cost which was a bit confusing. Their monthly spend went from $900 CAD to about $650. I never had a chance to dig deep on that, but I assume that before there was something clumsy happening with allowing expensive resources to live too long or something. It was a big difference for such a small company though.

I’m not sure if that’s a thing (Azure being cheaper) but it made me think I should do price comparisons there next time I have a budget conscious client who isn’t already partial to another platform.

Seriously, 10s of millions wasn't worth switching?

Thanks

I wager it’s just expensive to run a registry. I wouldn’t be surprised if this is what they’re really selling (the TLDs), rather than the registrar.

I suspect Google finds it a lot easier to run a registry with a few enterprise clients than to deal with flesh and blood customers owning domains.