Struct BlobStream 

pub struct BlobStream { /* private fields */ }

One opened blob, ready to serve ranges. Held by a host that is streaming or seeking a blob (playback probing a codec header, then a tail, then decoding forward) rather than loading it whole.

A range costs the bytes it returns, from either source, but for different reasons:

• Local (an external file, the local store, or a cache copy) — the stream holds the open file and every range is one positioned read of it. No hashing: a local file’s current bytes are the answer to a read of it, and a blob’s bytes are checked against the hash its row declares at publication, which is where they become canonical synced content.
• Remote, uncached — the stream fetches only the sealed chunks covering the range and opens them. A chunk that opens is authentic: the provider holds no key and cannot forge a tag, and the tag covers the chunk’s bytes, its index, and the header framing the blob. So verification is per chunk, which is what lets a range cost a range rather than the object.

Holding the local descriptor is a property, not an optimization: a path can be replaced between two reads, a descriptor cannot, so the stream keeps serving the file it opened even if that file is later evicted, renamed, or replaced. An in-place rewrite of that same file does reach the stream — that is a file the user owns and edits; coven’s own copies are published by rename or hard link and never written in place.

Implementations

impl BlobStream

pub fn plaintext_size(&self) -> u64

The blob’s whole plaintext length. Every range must lie inside it.

pub async fn read_at( &self, offset: u64, len: u64, ) -> Result<Vec<u8>, BlobCacheError>

Serve len plaintext bytes starting at offset.

len == 0 is an empty result, and an offset + len past the blob’s plaintext size (or an overflow) is an error, never a short read.