Comment by neixidbeksoxyd | Hacker News Reader

Comment by neixidbeksoxyd | Hacker News Reader

I think being part of a big company will always be like that. The 10x performers will get paid 1.3x and the .5 performers will get paid .8x, so if you're smart and not particularly passionate about your work it makes much more sense to be a .5x performer. Also, usually the codebases tend to be quite complex with little to no documentation since that kind of work is not rewarded. This means the hundreds of engineers who later work on that will take 10x as long to complete anything but it doesn't matter because the person who wrote it already got promoted and left the company. Then these engineers lose motivation as a task that should have taken 1 month takes 10 months, and that also brings productivity down. This is why I look for teams that use open source technology because there is competition there and if your documentation sucks no one will use your code or another competitor will take over. This is why I'm convinced places that use proprietary tech are going to fail in the future since the natural forces of competition will cause the internal technology to stagnate.

> This is why I'm convinced places that use proprietary tech are going to fail in the future since the natural forces of competition will cause the internal technology to stagnate

This was Bezos' reasoning for making as much of their internal software available to the public via AWS, a) now it's required to be documented, b) it's now subject to competition, you know it's not the best if no one's using it.

This meme needs to die. AWS was always built from the ground up to be a new product not an offshoot from Amazon retail.

https://www.networkworld.com/article/2891297/the-myth-about-...

Jeff issued a famous memo early on that mandaded public interfaces even for internal usage [1].

GP was talking how this improves reward structure, not about AWS

[1] https://medium.com/slingr/what-year-did-bezos-issue-the-api-...

Which has nothing to do with

This was Bezos' reasoning for making as much of their internal software available to the public via AWS

Those APIs had nothing to do with AWS. So how was the original poster not talking about AWS when they mentioned AWS?

Also a clean public interface says nothing about how badly written and documented the underlying code is.

I'm using AWS to mean the 212 different cloudy services they offer, not the original compute and storage products

No Amazon retail doesn’t develop a service internally and turn it over to AWS so they can sell it to customers. AWS treats Amazon Retail as just another customer that sometimes gets access to services before they are made publicly available. They talk about this all of the time at reinvent.

I mean Amazon as a whole, not specifically the retail arm. Wasn't the point that every internal service should be able to be productised? This was much later than when AWS was started.

So I guess I should disclose this now. I work as an implementation consultant (not my official title but it’s more descriptive - I actually do hands on keyboard coding) at AWS. But I’m just as far removed from the going ons between AWS and Amazon retail as anyone on the outside. So I’m definitely not trying to do the “appeal to authority”.

The mandate was in 2001. AWS first launched in 2006. From reading the letter, it wasn’t about being able to make services productizable. It was more so teams could work independently and choose whichever underlying technology they wanted. It also prevents the issue the original poster was having. It’s much easier to make changes to a small API than a monolith.

It takes a lot to go from internal API to product even on a small scale. Werner Vogels said at the last Reinvent that S3 is made you of hundred or more separate internal services exposed internally via an API. My last company we also had a mandate to be “API first”. Not because we were trying to be like Amazon with less than 75 people, but we actually sold access to our APIs to our customers - large business that used them as the backend for their websites and mobile apps. We also used the APIs internally for our websites and large ETL jobs where sour clients would send us files for bulk changes.

> This is why I'm convinced places that use proprietary tech are going to fail in the future since the natural forces of competition will cause the internal technology to stagnate.

I doubt proprietary tech is going to fade completely, but it's true that maybe there are forces that are pushing for big companies to make more and more projects open source. If you think about it it makes sense: the employees are happier because their work has more impact, they also benefit because they can showcase it in their CV. For the company they externalize testing, quality improves because employees are going to be more careful when their code is in the open and they are more motivated. It can also make hiring easier.

This is a very interesting perspective, had not noticed this advantage for companies to have more open source projects.

I think you're incorrect, although I hope you're not.

This idea will never become a reality for the vast majority of companies, for the exact same reason that companies fear piracy. If I open source my product then anyone who uses that open source code is a lost sale. A lost sale means I make less money, less money means the board is mad, on and on and on...

The bottom line is money, and open sourcing their code means (to them) a loss in potential revenue.

A counterpoint is that companies making money through sale of software code alone is the lowest ever. Google loses nothing from open source, as they sell nothing. The biggest companies have moved to SaaS stacks or enterprise sales models. We're seeing less value in the code itself, so the release of it is getting easier every day.

Well, that’s only true if you plan to sell the code. Much code exists not for its own sake, but to support the business. Maybe you need to build an in house inventory system for some kind of complex product you’re building. You were never gonna sell that software, but you still won’t open source it because that would help your competitors...

Well, and if it's so tied into an in-house inventory system just tossing a bunch of code into a Github repo is about 99% useless. There is almost zero value to just tossing some open source code over the wall.

Open source is going to be used to co-opt people into service contracts. For example one product I use right now. If you use it in a very particular way you do not have to pay much if at all to use it. But if you get out of that lane you need to goto the 'service contract' route. I am not talking 10 dollars a month either. I am talking 1-3k per machine. These things almost all want clusters. So usually at least 3. If you are using that level you probably will want at least 2 areas (dev, prod), more if you are doing it 'right'. So now your 'free' stuff just went sideways and now costs 200k+ just to get the software, per year. Oh but just use AWS/Azure/Google you say, add even more to that cost as they bury it in their usage fees. Then on top of that you need to develop your own programs.

I predict the open source bits will be bait. With many 90% solutions. The proprietary bits will be the ones you need to make it work like a real program. Oh there will still be soup to nuts full on free stacks. But I seeing more and more of this service fee way.

As for making hiring easier? Not so much. When you can get 100+ applicants for 1 position. The reality is at least 99 of those have to go away. One more filter does not do much other than let you round bin things faster say 'cant find anyone' then grab your favorite contracting firm and hire them anyway.

Elasticsearch was (is?) like that with AWS. The managed version lacked a lot of important enterprise features (LDAP integration) and was "optimized" to require several times as many machines for the same storage, since you had a ~1.5TB disk limit per node.

But I went from needing technical support from elastic 4-5 times a year, to zero and found work arounds for the other limitations, like cognito for authorization and storing less data in the cluster. The end result was a six-figure savings on licensing and less weekend work for me at the cost of a five-figure increase in AWS costs.

I'd say it was worth it.

Reply on news.ycombinator.com