Why is it difficult to write about advanced information technologies?

causa-arcana.com

5 min

View Original

Once we published an article “Decentralized services versus distributed ones”[1], which attempted to encourage readers to think more thoughtfully about terms such as “centralized,” “decentralized,” and “distributed.” This is important not only from the point of view of theoretical computer science and the convenience of practical activities in writing software. Often such terms are used speculatively for the purpose of advertising promotion of products of dubious quality, such as some implementations of decentralized finance (DeFi; from Englishdecentralized finance), or even completely fraudulent offers.

We did not intend to develop this topic, since any attempt to draw up a detailed classification is doomed to failure. Technologies are developing faster than they can be comprehended, and their complexity does not provide a chance to cover all the subtleties. The only thing that can be done is to advise you to study in more detail what you are going to use.

It is natural that we also did not want to respond to criticism of the draft classification we presented, because it would most likely be fair. Any choice of terminology has both advantages, allowing you to more conveniently convey thoughts, and disadvantages, shackling consciousness into too strict frameworks. At the beginning of our article it was said that our goal is not to create an ideal classification.

However, we were faced, among other things, with accusations of ignorance. This may mean that we have explained our motivation too poorly. We want to correct this error because the problems of epistemology and language are quite important to us, they concern thinking about the world in general, and not just about technology.

The same “wrong” picture

What's wrong with the common uses of the terms mentioned? Let's start by looking at the picture presented above, an alternative to which was shown in our article. It is often used to illustrate relevant terms, not only in popular articles, but also in scientific publications. Thus, in one of the publications we reviewed, a decentralized system is defined, among other things, as a system without a single point of failure, and this picture is immediately given.[3] If If you take a closer look at it, in the example of a decentralized system there is one node, when removed, the parts of the system will no longer be connected. Perhaps this is not a contradiction in some exotic definition of system requirements, the meaning of nodes and connections between them, but the publication in question does not contain anything like that.

The picture above comes from the 1964 book “On Distributed Communications Networks” (by Paul Baran).< a i=3>[4] It defines a decentralized network as one where the failure of a central node does not lead to a complete loss of connectivity between all nodes, but only some. This was quite relevant at that time, since the amount of information transferred between remote nodes of large networks was small, and connectivity within organizations could be much more important than between them. The same cannot be said about our time. In addition, the author considers a rather narrow issue of network topologies, that is, physical connectivity. Modern systems have many other properties, such as trust between nodes or storing data on different nodes.

Classification of network topologies. Possibly also incorrect or incomplete

An attempt to create a modern classification was made, for example, in the 2020 work “Decentralized vs. Distributed Organization: Blockchain, Machine Learning and the Future of the Digital Platform»[5] There it is proposed to perceive decentralization as dispersion of communication, and distribution as dispersion of decision making. This is an interesting work with a historical overview, its own theoretical apparatus and large-scale conclusions, including those of a socio-political nature. However, its results contradict established terminology and are also too complex to be applied correctly. So, in addition to the existing three terms, another one is introduced there - concentrated systems. Authors of texts cannot cope with three.

A similar approach is proposed by the creator of the Ethereum cryptocurrency Vitalik Buterin in the article “The Meaning of Decentralization”. [6]He especially emphasizes that each system can be classified in one way or another according to each of three different criteria:

• Architectural decentralization - how many nodes are there in the system? How many of them can fail without causing the entire system to fail?
• Political decentralization - how many people or organizations control the system?
• Logical decentralization - are the interfaces and data structures of the system a single monolithic entity (as in a blockchain) or not (as in federated social networks)?

We agree that whether a system meets any criteria depends on the goals set by those discussing it, but explicitly introducing such criteria again complicates the classification too much, leaving it potentially incomplete.

The main objection to our article was the contradiction (only apparent) of our definition of a distributed system with the most common one, according to which it is a collection of nodes that appears to users as a single coherent system. This definition is given by Andrew Tanenbaum himself in the book “Distributed Systems: Principles and Paradigms”. Our definition may be broader, but it certainly fully includes what is indicated. There are, for example, distributed databases, which are managed by one authority and make some administrative decisions. However, the need for an administrator in the case of databases is beyond doubt and may not be mentioned during the discussion. Most of the time, they operate autonomously, without any single decision-making center, and decision-making is actually carried out by the entire network using distributed consensus algorithms such as Paxos or Raft.

Thus, the convenience of classifications and definitions strongly depends on what problems need to be solved. As we already wrote in the article“Why is the word Internet written with a capital letter?”[2], we do not tell you how or what to say, we only help you think more effectively and avoid falling into mental traps set by those who really want to control you. This is not an easy task in such a fast-changing world, but with your support and feedback we can do it more effectively.