Chris @50htz@cathode.church

@stman @thatguyoverthere @50htz @vidak

I had to refresh my knowledge by reading the Wikipedia articles. Mainly, it came to my attention that the OSI model came with a standard set of protocols that lost the Protocol Wars with the TCP/IP suite. Every layer had multiple protocols, with many optional features that ended up being too complex to implement sufficiently -- the ISO group did not design OSI with ease of implementation in mind. I am fairly certain I have never used any of these OSI protocols in my life.

who is the biggest (most complex to implement) cat on the pic?

Just a wild guess, but I would expect the bottom three layers to be the most complex to implement. Maybe network layer is most complex since it has to do with routing across the planet. I think what we see today is that the application layer becomes the most complex -- again, more of a wild guess. But now that the Internet is a given, people want to make their own protocols on top, and network bandwidth and latency is good enough that the speed can hide the layers of indirection going on.

Another similar question: Is complexity distributed equally between all layers? Why?

I don't know. Maybe they had this in mind when they designed OSI. Assuming your referring to the Internet Protocol suite and what's we have on top instead, I think the answer is no. While there's a lot of nonsense like TLS and BGP, I assume it's nothing compared to the diversity of protocols at the application layer. Like for example, every multiplayer video game is implementing some bespoke protocol layered on top of TCP or UDP.

And is such distribution dependent of something? If yes, what is it?

The way I see it, networked computer architectures have been designed by accretion: by adopting what existing technologies seem to work for the widest variety of use cases, in order to support marketization and productization. There must be some exceptions but I am not that familiar with all the history. Was TCP designed for elegance? I don't know. The design space of computer technologies is so huge and we see such a diversity while at the same time they often overlap and have significant deficiencies such as lack of encryption. But for example, why did Windows not support ZIP files natively for the longest time? They wanted to support the marketization of compression technologies and not take sides? Or for media files they pushed their own proprietary technologies rather than support an open specification. Complexity sells. What incentive now does Microsoft have to make cybersecurity less of a problem when there is a whole industry popped up? Similarly, what incentive would there to be to develop and support decentralized Internet identity technologies when they can engender a market where companies spring up to offer their various solutions built atop everything else?

Is the layering of this model dependent of something(s)?

Presumably, it appealed to the intellects of the designers, even though it is the opinion of some IETF developers that "layering [is] considered harmful." This RFC is very interesting by the way, and I will want to read more: "Some Internet Architectural Guidelines and Philosophy." They talk a lot about complexity and simplicity. In my systems architecture class we learn that layered architectures can be good, since it separates concerns and allows for modularity and upgradeability. But this view assumes that the layers have loose coupling. For the Internet, the layers seem to be tightly coupled: "As a result of inter-layer dependencies, increased layering can quickly lead to violation of the Simplicity Principle. Industry experience has taught us that increased layering frequently increases complexity and hence leads to increases in [operational expenditures], as is predicted by the Simplicity Principle."

What is the logic, concept, if any, of this model?

It is a taxonomy that appeals to a kind of system designer. It is imagining that the layers can be separated and stacked in this way, but it ignores the realities. You could say that it was a mistake in drawing system boundaries. Again, I believe they designed things to make opportunities for corporations rather than make something practical that is easier to understand in detail.

[What] about a design constraint in protocols stacks / layers where complexity would be mandatorily distributed equally across all layers? What consequences of such constraint?

It's an interesting question. I'd like to think on it some more.

Enjoy this clip from Computerworld, where they refer to the complexity monster: Frank, Ronald A. (1975-10-22). "Battle for Access Standards Has Two Sides". Computerworld. IDG Enterprise: 17–18.

Does anyone know a reliable Minecraft to Minetest world converter? I tried a couple of Python ones and they weren't great.

I ask since I have a Minecraft world I'd love to keep, but I am not thrilled about having to set up a M$ account to keep playing.