streams are too high level

I would like to suggest making the base level operation a simple read of a byte range that returns that exact byte range, and then build the stream api on top of that.

basically, just something analgous to POSIX _read_ and _write_ https://www.gnu.org/software/libc/manual/html_node/I_002fO-Primitives.html

if you are building an interesting on disk data structure like say, an append only b-tree, there are a lot of small things to read (branches in the tree) these are likely to be a single block or smaller, so creating a whole stream for that is a significant overkill.

streams can be useful too, but if you have a simple read/write you can _make_ efficient streams, but not the other way around.

The issue names no files or tests. Start by locating the current stream API and its read/write entry points, then compare them with the proposed exact byte-range operations. Done means agreeing on the base API and how streams would be built on top of it.

Written by the indexing model from the issue text.