Re: [fuse-devel] FUSE i18n

[Akira Y. <yo...@de...>]

Hi,

On Thu, 08 Dec 2005 12:24:21 +0100
Miklos Szeredi <mi...@sz...> wrote:

> Hello,
> 
> > I made a patch to add encoding conversion of file/directory names to FUSE.
> > It will be useful various filesystems based FUSE, especially for persons
> > without english. I attached my patch. Please try it.
> 
> Thank you very much for your contribution.  I have a few comments:
> 
> Would it not be better to perform the conversion in the low-level
> interface?  That would affect much fewer operations (12 instead of
> 30), and would have better performance.
> 
> Also I'd very much like to have a generic stacking interface for FUSE
> (both for the low and high level libraries), so that other features
> like this may be added in the future, without needing to modify the
> core code.  The interface is not yet clear in my mind, and I'd be
> happy to have a discussion about this.

I have considered your comments, and now I think that my current approach
is one of the best implementation. There are two reasons.

a) lowlevel operations encapsulate requests and file/directory name
infomation is in them. If we do file/directory name conversion in lowlevel
operations, we will need to rewrite request structure. I don't think that
is so beautiful.

b) it's a stacking structure like you said, and a good way for minimum
change. For example, see below:

- the original FUSE path
 fuse_ll_process() -> do_open() -> fuse_open() -> f->op.open()

- FUSE i18n path disabled i18n feature(without bcode/fcode options)
 fuse_ll_process() -> do_open() -> fuse_open() -> f->op.open()
	(same as the original. no change)

- FUSE i18n path enabled i18n feature(with bcode/fcode options)
 fuse_ll_process() -> do_open() -> fuse_open() ->F->op.open()-> f->op.open()
	(F->op = FUSE i18n operations: API compatible to f->op)

Welcome comments. Thank you.

Best regards,
	A.YOSHIYAMA <yo...@de...>

p.s. I want a mutex structure for struct fuse_operations.