FileCore write support in Linux kernel

This is an unusual suggestion as it doesn’t actually run on RISC OS, but bear with me…

Linux’s ‘ADFS’ module can read FileCore partitions. That means you can mount an ADFS, IDEFS, SDFS, whatever disc and read the RISC OS data directly out of it. An optional flag ‘ftsuffix=1’ allows filetype information to be appended to filenames in the same way as NFS – so you could plug in a RISC OS disc to your NAS and Sunfish would see the right filetypes.

Linux has ‘experimental’ FileCore write support. However it cannot allocate blocks, because it doesn’t handle writing to the FileCore free space map. So it can write to F format discs if you don’t change the length of a file. It can’t write to F+ discs at all because the catalogue on F+ is variable length (due to long filenames) – changing the length would involves claiming or freeing blocks, which it can’t handle.

Finishing off Linux FileCore write support (and testing it to the point where it’s reliable) would be a key driver for a chunk of backend automation regarding automatically building disc images. Longer term, it could also mean that FileCore became a ‘first class’ partition format under Linux, so you could easily plug in a RISC OS SD card or USB stick into Linux and it would just work.

Linux kernel development is a slightly different set of skills to those normally required here, but it also requires a willingness to dive into the detailed FileCore data structures.

One thing, I think there was discontent about Filecore as it is, and any changes might have to be done there first, or with Linux compatibility in mind. Maybe you could work in reverse and have a RISC OS partition that Linux can create and use, and change RISC OS to adhere to the new format. G for GNU (-: