[DGD] Buffered I/O in atomic mode

[Mikael Lind]

Shentino,

On 25/09/2007, Shentino <shentino at gmail.com> wrote:
> Just htought of a neat way to get around the "no fs changes in
> atomic"...just buffer the writes.

Yes. Do you intend to reflect those writes in subsequent reads from
that same atomic execution round?

> However, things could get real messy if I attempt to rename files.

Messy in what way? It seems like a challenge to get this right, at
least if you want to reflect all operations in subsequent operations.
You would perhaps need to build a file system buffer in the mud
library and mask all file access efuns in your second-level auto
object if you are using the kernel library.

> So far, I'm thnking of taking inspiration from the Linux kernel's
> dcache and icache.
>
> Is supporting filesystem reorganizations in atomic mode worht hte trouble?

Could be. Are you trying to solve a particular problem or are you
doing this for fun?

Mikael