So we were stuck with a product which we were writing against something that the official word was would no longer be supported by the time we launched, while being told to develop against the platform which is "beta" and they don't want to commit to supporting the feature set of - and which plain couldn't be used yet at the time they told us this.
This is just a ridiculous way to run a commercial platform offering (aka: why I always tell people to use boring VMs for as long as possible).
Shouldn’t it be the other way around? Take whatever nice features you can “for free” as tech debt, and if that api inevitably disappears… push it down the stack and own a bit more of it.
If you have to build everything from first principles it’ll take way longer to get going.
I neither support nor deny this belief but it's an increasingly common impression.
You're right, it is FUD, but it's not malicious FUD spread by us to spite google, it is actually rational FUD based on real reports that is entirely google's fault
I guess I'm lucky that I haven't pissed off the GCP algo yet.
If you see risk as binary, then yes there is risk in any SaaS provider, or indeed any part of your company supply chain. Risk though is not binary, it is measured on a scale of 0 to 1.
There is risk every time you get in a car. But some cars are safer than others. Some are renowned for putting safety first.
The vast majority of people survived their Ford Pinto, no doubt some loved it, but the perceived risk of driving it (rightly or wrongly) was higher than say a Volvo.
Google is the Pinto of SaaS. Whatever the actual risks are, they are perceived as being higher risk than other SaaS providers. Thanks to their "no support" policy, the penalty for failure is total extinction. With most SaaS businesses there are humans in the loop who can make human decisions.
That's some strong personal opinion... source?
GCP has been extremely performant and reliable for me across several different companies.
I'm not saying that people have definitely run into issues, which we've all read about here on HN, but this sounds more like squeaky wheel than the norm.
> Thanks to their "no support" policy,
Their support has been excellent, when I've needed it, which is rare, since it has just worked for me very well for a decade now. Their documentation is also pretty well done too. Just like with any sort of SaaS solution, you should be building a relationship with an account exec.
I've had CloudFlare start to put in weird restrictions on my account once I hit a certain size. It showed up with requests being oddly denied and zero notification. I contacted my account manager (called their cell phone!) and the problem was resolved in a few hours. I don't even pay for a business plan, but I did make sure to develop a friendly relationship with them when they originally reached out to me.
It's absolutely personal perception. Ford made a couple millions Pintos, I never had one, much less saw one explode in a fireball, but I'm not rushing out to buy one either.
My perception is borne from a long history of reading stories here. Over time they create a perception in me that Google regularly drops products, changes APIs, changes pricing, closes accounts, withholds earnings and so on.
I'm sure millions of people happily use their services. However my perception of risk with Google is high, so as a result I don't use any (paid) Google services and I take my business elsewhere. Obviously I use Google search, and watch YouTube, but I'm happy to not put Google in my supply-chain, nor rely on them for revenue.
Now maybe it's just bad PR. Maybe Google has real lower risk than say someone else. Maybe Pintos were statistically safer than a Volvo. But perception is everything, and my perception is that, given a choice, I'm not going to use GCP.
If what you are doing is important get a sales rep and invoice billing. This is actually tru of any service, but Google is particularly bad. You don't want to be caught up in their automated fight against fraud and abuse
I figure Google knows how to protect against attacks way better than a random email provider. I enrolled in their Advanced Protection Program [1] because I don't want my email taken over. I've used a Google Voice number deliberately because somebody can't walk into Verizon or AT&T and get my number.
The idea that there's absolutely no recourse if Google decides I'm a bad guy. Especially since they sometimes seem to target associated accounts - are all of the family accounts going to be cancelled, too?
I don't think Apple's security is as good but I'm also not qualified to judge that. And at least I can talk to support on the phone or go into a store. So I'm moving my email that way over time.
[0]https://www.theguardian.com/technology/2022/aug/22/google-cs... [1]https://landing.google.com/advancedprotection/
If the way Google has decided you should build something happens to work for you, great, but for most people, they want a product built around how they like to operate, not get told how to do it.
But also Google has a trust issue. I think GCP makes a superior product, but I would never use it. I'd be afraid that one of my former employees leaves the company, does something Google doesn't like, and they shut down my entire GCP account because my gmail is associated with the gmail of someone who did something bad, even though they don't work for me anymore. (Yes, this is a real thing that happened)
I can't trust Google not to just shut down my account and then give me no human to talk to to get it fixed.
> I think GCP makes a superior product
Exactly. They do.
The rest is mostly just conjecture. The loudest wheels get the most notice... there are plenty of other people, like myself, who have been using GCP for a long time without any drama. It just works.
AppEngine had a lot of limitations, like the version of Java you could use, because they had to basically hack at the JVM to get it secure enough.
These days though, things like Cloud Functions, are effectively just simple containers and http endpoints. I could move them off to another provider with a days work.
Quickly convinced my friend Jeff to get onto it... we saw a need for a datastore wrapper and he (and I a tiny bit) wrote Objectify, which is one of the most widely used tools for AppEngine out there.
Look at the license.txt... Feb 6, 2010, 13 years old...