Only `.zip` and `.sqfs` archives are supported as these are the only
common archiving formats which make it easy to read directory listings
from the archive without downloading the whole archive.
Internally the archive backend uses the VFS to access files. It isn't
possible to configure the internal VFS yet which might be useful.
## Archive Formats
Here's a table rating common archive formats on their Cloud
Optimization which is based on their ability to access a single file
without reading the entire archive.
This capability depends on whether the format has a central **index**
(or "table of contents") that a program can read first to find the
exact location of a specific file.
| Format | Extensions | Cloud Optimized | Explanation |
| :--- | :--- | :--- | :--- |
| **ZIP** | `.zip` | **Excellent** | **Zip files have an index** (the "central directory") stored at the *end* of the file. A program can seek to the end, read the index to find a file's location and size, and then seek directly to that file's data to extract it. |
| **SquashFS** | `.squashfs`, `.sqfs`, `.sfs` | **Excellent** | This is a compressed read-only *filesystem image*, not just an archive. It is **specifically designed for random access**. It uses metadata and index tables to allow the system to find and decompress individual files or data blocks on demand. |
| **ISO Image** | `.iso` | **Excellent** | Like SquashFS, this is a *filesystem image* (for optical media). It contains a filesystem (like ISO 9660 or UDF) with a **table of contents at a known location**, allowing for direct access to any file without reading the whole disk. |
| **RAR** | `.rar` | **Good** | RAR supports "non-solid" and "solid" modes. In the common **non-solid** mode, files are compressed separately, and an index allows for easy single-file extraction (like ZIP). In "solid" mode, this rating would be "Very Poor." |
| **7z** | `.7z` | **Poor** | By default, 7z uses "solid" archives to maximize compression. This compresses files as one continuous stream. To extract a file from the middle, all preceding files must be decompressed first. (If explicitly created as "non-solid," its rating would be "Excellent"). |
| **tar** | `.tar` | **Poor** | "Tape Archive" is a *streaming* format with **no central index**. To find a file, you must read the archive from the beginning, checking each file header one by one until you find the one you want. This is slow but doesn't require decompressing data. |
| **Gzipped Tar** | `.tar.gz`, `.tgz` | **Very Poor** | This is a `tar` file (already "Poor") compressed with `gzip` as a **single, non-seekable stream**. You cannot seek. To get *any* file, you must decompress the *entire* archive from the beginning up to that file. |
| **Bzipped/XZ Tar** | `.tar.bz2`, `.tar.xz` | **Very Poor** | This is the same principle as `tar.gz`. The entire archive is one large compressed block, making random access impossible. |
## Ideas for improvements
It would be possible to add ISO support fairly easily as the library we use ([go-diskfs](https://github.com/diskfs/go-diskfs/)) supports it. We could also add `ext4` and `fat32` the same way, however in my experience these are not very common as files so probably not worth it. Go-diskfs can also read partitions which we could potentially take advantage of.
It would be possible to add write support, but this would only be for creating new archives, not for updating existing archives.
<!-- autogenerated options start - DO NOT EDIT - instead edit fs.RegInfo in backend/archive/archive.go and run make backenddocs to verify --><!-- markdownlint-disable-line line-length -->
