Live data from Hacker News

Ask HN: What is the most complex software that you have built/worked on?

news.ycombinator.com

1–10 of 43 posts

Re: Ask HN: What is the most complex software that you have built/worked on?

#3
During 1978-1979 single-handedly writing all of the data processing programs necessary to automate all departments in a 150-employee manufacturer of computer terminals on a 32KB RAM 50MB HD multiuser Pick O/S Microdata: order entry, purchasing, shipping, receiving, payroll, accounts receivable, accounts payable, MRP, QA, work orders, etc, etc, etc.

Re: Ask HN: What is the most complex software that you have built/worked on?

#5

During 1978-1979 single-handedly writing all of the data processing programs necessary to automate all departments in a 150-employee manufacturer of computer terminals on a 32KB RAM 50MB HD multiuser Pick O/S Microdata: order entry, purchasing, shipping, receiving, payroll, accounts receivable, accounts payable, MRP, QA, work orders, etc, etc, etc.

That sounds like fun

Re: Ask HN: What is the most complex software that you have built/worked on?

#7
Hmm. A lot of software is unnecessarily complicated. This has to do with "too many cooks" or "design by committee". The term "complex" is relative. Is a trading system complex on its own merit or is it the process of building it that is complex? I haven't worked on a major project so I can't say.

I took a programming course many years ago. Not because I needed to learn programming but to get job interviews. The instructor asked us to write some C code to perform some small task. I completed the coding in about 5 minutes and it was just two lines (my prior experience was 6502 assembly not C). Others in the class were writing screenfuls of code. I imagine that the same situation occurs on large projects where mediocre programmers write more code than required for the task. Projects get bloated and more complicated as a result.

Another anecdote. I wrote a small operating system for a piece of hardware known as "touch memory" made by a company called Dallas Semiconductor. Intriguing little device, I must say. This was 20 years ago. My boss met up with the owner of a company that was attempting to build what I had already built in just two months. His 4 programmers could not accomplish much of anything over the course of an entire year. I guess that means that I was 24 times more effective. The company went bankrupt.

The most complex project I worked on? I dunno. Maybe converting PDF into braille. Would you say that is complex? I wrote all of the code from scratch without using any third-party tools. That's another area where projects go wrong. They rely too much on closed-source middleware that their own programmers don't understand. It's a recipe for disaster.

Re: Ask HN: What is the most complex software that you have built/worked on?

#8
Well im about to release the alpha but basicly its a project i worked on ~4 years in privat time and consists of two parts.

1. A custom in-memory graph storage/database which is threadsafe and designed for fast multithreading purposes. It also comes with a custom query builder/language which can be transportet via json so viable for every language. It can either be used by directly importing it as dependency in your go project or build as server with a http api package i build and used via api.

2. An architecture/framework which enables you to create completly self supervising software that basicly only persist of you defining a set of "abilities" which you do in form of go plugins. A plugin defines the parameter structure it needs to get executed. The architecture uses the in (1) described graph backend to store all the data. If data gets mapped into the storage the architecture will effective check if the new data can supply any of the registered plugins, and if so execute it. Since the architecture runs multithread by default all jobs will be parallelised into worker/runner threads. Results from those will be automaticly checked by a scheduler to see if it can supply new jobs with the new knowledge (and optional added already existing knowledge). Also data returned from a plugin will automaticly be mapped into the existing storage. Since all the runner do scheduling based on new data you have no central supervising but rather thread-distributed self supervising.

This way you basicly just define certain abilities as plugins and than insert the starting data. The architecture/framework will take on from there.

This was a strongly simplified explaination for what it does and i coded alot of different stuff in the last 15-20 years tho this project by far was the most complex in terms of dynamic data mapping / scheduling / etc. So complex in terms of logic rather then in size.

Re: Ask HN: What is the most complex software that you have built/worked on?

#10

During 1978-1979 single-handedly writing all of the data processing programs necessary to automate all departments in a 150-employee manufacturer of computer terminals on a 32KB RAM 50MB HD multiuser Pick O/S Microdata: order entry, purchasing, shipping, receiving, payroll, accounts receivable, accounts payable, MRP, QA, work orders, etc, etc, etc.

Did they use this up until the 2000's? It's the kind of software that sounds difficult to replace.
Post reply on HN