GSoC ( Ajay Rajera ): Support generating filesystems from manifests with mkfs.erofs

[Ajay]

Hello EROFS Maintainers and Mentors,

My name is Ajay Rajera, and I am here to express my interest in
participating in Google Summer of Code with EROFS, specifically for
the project: "Support generating filesystems from manifests with
mkfs.erofs" mentored by Chengyu Zhu and Gao Xiang.

Over the past few weeks, I have been actively exploring and
contributing to the EROFS ecosystem to familiarize myself with the
codebase. Some of my recent contributions include:

1. Addressing the excessive default verbosity in mkfs.erofs by
suppressing per-file processing messages (Issue #14 in erofs-utils). I
actually did it but couldn't submit the pr as it is restricted there.
2. Improving and updating documentation within erofs-utils to fix
grammatical errors and outdated information. (
https://github.com/erofs/docs/pull/25#issue-4078189227 )
3. Implementing support for EROFS device tables in the go-erofs
library, which involved parsing device tables from the superblock and
mapping block addresses. (
https://github.com/erofs/go-erofs/pull/18#issue-4078327809 )
I am looking forward to contributing more and keep learning from it.
I am drawn to this GSoC project because I have experience with file
system concepts and C programming, and I recognize the limitations of
currently generating images strictly from source directories or
tarballs. Implementing support for manifest files (like
composefs-dump(5), modified unix proto files, or BSD mtree(5)) will be
a highly valuable addition, providing better metadata control,
ordering control, and better performance for large filesystem trees.

I have started drafting my proposal so I also have a few initial
questions regarding the project:

1. As I detail my implementation timeline and architecture in the
proposal, are there any specific edge cases or phases of the project
you recommend I focus on the most?
2. The project mentions supporting at least two common manifest
formats. Is there a preferred format (e.g., composefs-dump vs. BSD
mtree) that you would recommend I target first for the initial
prototype?
Thank you for your time and for maintaining such a great project. I
look forward to your guidance and the opportunity to contribute
further during GSoC.

Best regards,
Ajay Rajera