muhanga @ muhanga @programming.dev

Posts

0

Comments

37

Joined

3 yr. ago

Game mods and Advent Of Code did it for me.

I did a small RimWorld mod and a parser for NoManSky internal format.

Creating both of them was a blast. I had fun doing programming stuff again.

Advent Of Code allowed me to try different languages in a small bursts of the different problems. Somehow I really like this format.

From my personal consumer experience I would say twitch streams and steam demo fest are my two main sources of new games to "put a pin". And on twitch I mostly watch small channels (below 200 viewers), with couple of exceptions.

When your jvm vendor is IBM? I don't see other reasons. OpenJ9 had quite good polyglot vm Platform. I am not sure how well it is now compared to the Oracle GraalVM. As a Singular developer you don't really care about your jvm vendor. Maybe if you already have a lot of IBM specific tooling OpenJ9 will be a nice choice.

First, thank you for sharing your perspective.

one useful innovation by providing an alternative system for digital transactions over the web

From my perspective this is exactly proves my point. Blockchain and crypto are solving their own problems.

As for Wakfu. From what I was able to "research" from open sources on this topic, they were blacklisted because they didn't want to fix their payment processing system. Their payment processing was not working correctly, users complained to the PayPal and credit card payment processing companies and this got them blacklisted. And they did added crypto yes, but I don't see how much this actually saved them, as I can find an open financial reports for that.

What I can see is that they are fixed their payment system, somehow (as forums are still contains user woes for payments not processed). If you have more sources on this problem/solution with Wakfu, please share I would very much like to know more.

And also there is a lot of cases where you really don't need or want static typing. Static typing and type systems are great when they helping you but very bad when you are forced to fight them due to compiler problems or bad modeling.

In the end it is all an engineering problem: which amount of your budget you need to spend on proving programm correctness. Cost/benefit and all of that.

Static typing and unit tests don't make your codebases great, safe and supportable. Thinking and understanding your usecases, decomplecting problems and some future planning wins.

I am very very skeptic about this whole DORA org and approach and adoption. I have a greatest respect to the DevOps philosophy, but from DORA I only got a baseless/faceless metrics. Maybe it is also Google driving it that gives me additional rejection impulses. I am quite skeptic at what and how Google does and this colors my perception of all thing it produces.

Why can't there be a normal P2P project handling exchange of information and/or modern fiat in the same way (Something like Paypal, but transactions have no middleman)?

Firstly because money is a physical, cultural and social construct so it can't be changed on purely informational basis. Someone still need to share burden of proof and they want to be compensated for the labor. So until we get a StarTrek replicators (mean we remove need to spend money on basic need and survuval of whole human race) this is a state we are in.

In short blockchain and crypto don't solve any real world problem. It solves problem that it itself creates.

I can sell you amazing knife and it will fix the world hunger, but only if you can buy bread and sharpen the knife. This is crypto sell point in the nutshell.

Blockchain is little different as it solves the problem of provable chain of evidence, but it is not economically viable due costs needed to run it for organisations that require it. Any problem that blockchain can solve require that all information for this will be stored on blockchain. And physical object information is not stored on blockchain, so data input errors/malpractice is still the problem and this reduce blockchain effectiveness to the basically zero.

Dan Olson aka foldablehuman have an amazing series of video essays regarding all the crypto blockchain and web3 scam running around. I highly recommend them. It just a sea of information regarding current state of things with crypto/blockchain business.

If you are using BitBucket Cloud you can create pr rules to include people into Review based on files change. And then you can create a user for a bot to monitor those PRs using standard BB notification emails. Of course if there is not much PRs bot is Overkill and human will be enough.

You can always "just" create a static script that pulls repo check diff for files and email people if something is found. This way you don't link your solution to the git cloud offering.

This is why computers had a big reset button.

This really devolves into "good teams can deploy daily, can raise a small PRs and have small number of rework". And this is like... thank you, but it is obvious. If team is able to do this things constantly it is probably a good team.

DORA says that if your team is able to do same pattern (as they show) it will be "elite/good" team. This really smell like a cargo cult. And managers are already using DORA metrics as good/bad teams metric.

This is clear Goodhart's Law case: ""When a measure becomes a target, it ceases to be a good measure". So either DORA knowingly did nothing to protect against metric gaming or they didn't considered impact they will make. Neither of those is a good in my opinion.

So yeah I don't like DORA in it current iteration.

Same problem with "top 10%".

"DORA guys" came to our org in the past. And sing a song of "all successfully teams do that to, so you should too". One of the my question, that was left unanswered, was did they analyse negative scenarios to check if their suggestions actually works and add too the reducing cycle times and what not?

And most of the time my cycle time is more depends on number of meetings I need to attend through day than on anything even remotely related to the coding.

I understand what DORA tries to do, but what they achieve is just another cargo cult.

Somewhere someone probably does... But this piece of code really look like someone either tried to inline a bunch of calls or this is code generated object mapper from json or other nested model.

Nobody with a sane mind and serious attitude will use this code as a "real" code. (I still believe in people, despite all the evidence to the contrary I get every day)

As a fun bit though this taken some dedication.

You are absolutely right. At first I just wanted to add my favorite language to the bunch, but then I realised that this isn't really answering anything, because the use case matters most.

You can use any language to programm solution to any problem in any environment. And given that here we have many developers fixing many different problems we will end with just a collection of all possible languages and problem/solution permutation.

Language doesn't matter. Solution and solution logic matter. And most times we are using a Plain Human Language to crate a solution and then encode it.

Plain Old Human language. Remember comments? Remember moments when things get very complicated and docs and comments become your only help?

That mostly because none of the languages is the best. Some of them better in some places and worst in others.

For example: Java. Amazing library range, enterprise support and feature and community reach. Java also fail in shambles when you need a low level or guaranteed performance. Erlang. Robust distributed and fault tolerant. Now try to create something that is not network, agent oriented and should work locally only.

Every language has a niche. Look at javascript. JS is only exist because of it's niche. It wasn't good as a language, but it was the only one viable solution in it's niche.

Same with assembly. Nobody sane would use assembly if it wasn't that close to the metal.

There are time tested solution in every niche and it is wise to know why they still there and what drives them.

I learned this lesson through one of my optimization tasks. Speeding up programm by just reducing input data by 80% solved multiple problems. And real eye openner was an "article" about grep: https://lists.freebsd.org/pipermail/freebsd-current/2010-August/019310.html

Basically if you have less work you will do it faster. Since then my first question is always: Can we do less work?

Learn to talk to people and maintain connections. It is most invaluable skillet that will help you both carreer wise and professionally. The more people you know the better it is for your carreer. Learn to present yourself. Visibility matters very much, so it also good to "sell yourself" sometimes. There is really fine balance between making a sell and just bragging, people don't like second, but okay with first. Learn to teach other people and help them. Most troubleshooting experience I get now is from helping other people. They have a completely different way of doing code that I am (as a whole) and I am just getting this free xp by helping them and also adding one more trouble to my personal solution cupboard.

As for technology, pick what you like and master it, but also make a peeks at what is currently "in vogue". For example I really have no depth knowledge in the current frontend space, but I did take a passing looks (and build simple tutorial projects) with react, angular and dart. It didn't really required a much effort from me, but this helped in the long run to be aware.

Just a note by setting up a 30 hours home project you effectively removing "people with lives" from your hiring pool. People who can do a 30 hours either have a lot of freetime currently, or code after the job. And if you really want those people in the team then go ahead, but you are missing on 8 to 5 crowd and that is a very good and diverse talent pool. From my experience 8 to 5 minded people are very good in solving tasks in sustainable manner. They just don't have time to fuck with the system and doing effective "dont-call-me-at-night" solutions.

if you are doing a lot of interviews you need a common set of questions and measures and this take a lot of time and effort to setup.

Personally I would suggest to setup interview as a two parter first ask some theoretical questions and then ask to create a simple code with simple problem related to the questions. This helps to find out if people are really understand what they talking about. This again require a lot of thought to setup an to have small practical tasks relevant to the questions.

For example in most recent interview I asked candidate about algorithm complexity, data structures, garbage collection and then asked them create a simple dictionary to store a hierarchical structure. This helped to see if candidate knows what he is talking about and can use his knowledge in practise. I have seen a lot of people without good theoretical knowledge, but they create a code that is good and working despite their gaps and other way is also correct people have a good theoretical knowledge but fails to apply it in practice.

So figure out who you are searching for. Create an ideal checking solution for their skills and start combing the desert. There is no shortcuts in hiring, sadly.