[haiku-development] Re: GSoC CIFS implementation
- From: Ingo Weinhold <ingo_weinhold@xxxxxx>
- To: haiku-development@xxxxxxxxxxxxx
- Date: Wed, 25 Mar 2009 00:36:29 +0100
On 2009-03-24 at 23:15:26 [+0100], Stephan Assmus <superstippi@xxxxxx> wrote:
> On 2009-03-24 at 22:56:14 [+0100], Obaro Ogbo <obaro.ogbo@xxxxxxxxx> wrote:
> > I've been studying the CIFS implementation idea and have a few questions.
> > Should it be written from scratch, or is a port/modification of an
> > existing one preferred?
> > If a port, linux or one of the bsd's?
> > Also, I've been trying to make sense of userlandfs. Any info about
> > userlandfs and links to general haiku filesystem addon information will
> > be appreciated.
>
> A couple of days ago, I did some preliminary research on this topic.
> Apparently, there are a few CIFS client implementations for FUSE, at least
> three, IIRC. All of them seemed to be using a library called "cifsclient"
> or something similar which is part of the SAMBA software suit. If porting
> the whole SAMBA package is too much work, maybe porting just that library
> is less involved. Then one of those FUSE implementations for CIFS could be
> ported. I am guessing that Ingo will have the FUSE bridge performing well
> in short time. :-)
Well, I think "working" well should be possible; "performing" is another
matter. The userlandfs adds quite a bit of overhead and the FUSE interface
with its annoying use of paths instead of nodes doesn't really make things
better. So an implementation with an Haiku FS interface is definitely
preferred.
Also, unless it has tons of nasty dependencies that have to be ported as
well, I'd be surprised, if porting the required samba parts was a lot of
work. So getting one of the FUSE CIFS implementations to work in Haiku might
turn out to be only a matter of days. Even writing a Haiku FS interface for
ported code is probably not enough to fill three months of project time.
CU, Ingo
Other related posts:
