夜の抜ける

Sometimes people confuses a complex design with a powerful one. The only relation I see between power and complexity is that both can be dangerous.

Everyone will agree that complexity is undesirable, and power may or may not be, depending on the context.

I tend to make things simple, specially for the programmer, because we’re the laziest creatures on the surface of earth (and perhaps many other planets). We just try to look productive from time to time and hope other races don’t notice.

We want something easy and intuitive to use, don’t we?

KISS FTW!

This happened about a year ago, but I forgot about it until now… I found this “feature” while developing a simulator to use with honeyd.

After I finished it, my employer said it should run on Windows too, since I didn’t used anything unportable, it should had to run without modifications, but it didn’t.

For some weird reason, undef evaluates to true on Windows… that’s a bug, undef is expected to evaluate to false always, it’s not an undefined beviahour. I can’t explain how a bug like that ended there, and I don’t want to know, because it’s a weird platform anyway.

Most “low-level” RPC frameworks are broken, badly written and unportable. I got a project to develop a middleware to help in the abstraction of RPC transports, but I ended implementing a new framework. Soon I will publish it on Gitorious :).

I’m posting this because everyone’s talking about Apple’s GCD. Which is a similar technology to the one included in my OS, however I’ve been developing it since Dec 2008, so there’s no relationship.

Apple got the wrong idea, mine is far better :P. Now, let me explain it:

Until recently my OS was just a bunch of papers and code snippets, but is progressing now :). It’s something completely different from any other OS, and while it will retain some compatibility (i.e. you will be able to run POSIX code on it), it’s designed to work in a very different way.

It aims to be an efficient OS, focused on desktop systems, with security in mind. Being desktop-focused just means it will be optimized for that role, but tunning it a bit you could make perform well on embedded hardware as well as giant NUMA systems.

It’s main objective is security. To achieve this goal I had to slightly redesign a lot of concepts which make an OS. One of those concepts is the API.

Since I’m trying to achieve as much security as possible (when it seems reasonable), it must enforce the concept of lowest privilege whenever possible. To make this feasible, applications are divided into modules, and while this division is explicit, it must be simple enough to make it the standard.

The solution is the development of a “components” concept. It requires a new executable format, but lets you easily divide programs into small chunks that could run in parallel (on-demand). This is where the similarity with Apple’s GCD ends.

The aforementioned “components” are completely independent, and not even share the same address space. They communicate through a lightweight IPC API which provides message queues, asynchronous RPC, transactions, and shared memory.

Another interesting feature is that it’s designed to be completely asynchronous.

Sadly I have not much time as I would wish to work on it…

Ohh, brave developers, keep away of privative software.
The darkness lies beneath.

Follow the path, THE PATH OF WISDOM!

I must hold the pace, no one will stop me.
Follow me brave developer, the truth shall enlighten you.
Follow the path, THE PATH OF FREEDOM!.

(PATH OF WISDOM)
(PATH OF FREEDOM)
(…)

[Please contribute, this song is under CC-BY-SA, it must be a hit!]

Please tell me that you didn’t!!

NO, we don’t need more GNU/Linux distributions! Not if they don’t provide anything new. If the features you want aren’t something that deals with the lower levels of the system, specially file hierarchy and package management, do not do it, it’s better to simply improve an existing distribution without forking it, or provide an “overlay” set of packages to customize it if the changes couldn’t be accepted otherwise.

There are thousands of non-innovative new distributions that provide nothing but chaos to the free software ecosystem, we don’t need more. It looks like everyone wants to do a new one.

And no, it doesn’t enhace your karma, it will be set to LLONG_MIN premanently, unless, you have a very very very goood reason to start a new distribution.

There are two special kind of distributions that are specially dangerous:

The politically-oriented ones, thus politically-discriminative, segregate the free software user community needlessly. This should be avoided by all means for obvious reasons, specially because that kind of politics have nothing to do with software, so don’t impose any kind of obstacle, everyone can work together.

Those that immitate other operating system look&feel too closely and strictly, don’t make any good since their objectives aren’t to improve the system, but just copy and promote some way of doing things, making the users actually believe there’s no diference, so implicitly promoting the other OS.

Also this leads to the users to think both should work in the same way and actually are the same thing, thus promotes a way of thinking and “culture”. This is specially annoying when it’s about Microsoft Windows way of “thinking” and “culture”.

Feliz cumpleeee!!!