notesthoughts

When something breaks, go one layer down

Learning about inodes made me consider what familiar tools hide. Everyday usage is often enough, but unexpected behavior needs another explanation.

While learning what an inode is, I became interested in the separation between a file’s name and the structure that manages it. In daily use, I tend to think of the name as the file itself. But the name alone is not enough to explain an open file or what happens after deletion.

Similar moments arise with other tools. Ordinarily, I start containers with Docker and update state in React. If work proceeds well through those interfaces, there is no reason to inspect the internals every time. But when something behaves unexpectedly, repeating the same usage steps is not enough.

When I need to know what was hidden

Abstractions let us avoid dealing with complicated details every time. They allow us to do a great deal at a higher level. When a problem underneath reaches the interface above, though, I need to step down briefly and look for an explanation.

I find this interesting. Being good at using a tool need not mean knowing all its internals. But it does require noticing when the current explanation is insufficient and judging which layer to examine next.

The question that began with inodes eventually led to design. Choosing not to add an abstraction is different from having no structure. Rather than adding layers for every possibility, I would prefer boundaries around the places likely to change in practice.

When adding a structure, I want to ask both what it can hide and where I can look if something goes wrong later. If it becomes easier to use but harder to explain or trace, that cost matters too.

I cannot study every tool’s internals before using it. I work through familiar interfaces most of the time and go one layer down when the explanation falls short. For now, that back-and-forth feels like a realistic way to learn.