samba - Nov 2009 - samba & libiconv

Hi,

I am building a NAS that only has a couple of SATA disks attached (no USB, no
CD/DVD, no NFS, no ...). Since the available space for the rootfs is limited
(only 8 MB), I'd like to omit libiconv because this library takes 875
Kbytes.

The filesystem on the disks will be ext3. UClibc is built with WCHAR support.
Can I build Samba 3.3.8 without libiconv and still have unicode filenames
(russian, english, ...) displayed correctly in Windows and Linux clients?

thanks,
theHog

----- Oorspronkelijk bericht -----
From: Jeremy Allison <jra at samba.org>
> On Tue, Nov 03, 2009 at 11:01:24PM +0100, Volker Lendecke wrote:
> > On Tue, Nov 03, 2009 at 01:43:42PM -0800, Jeremy Allison wrote:
> > > No. Samba needs libiconv to do the utf8 <--> UCS2
translations for
> > > clients.
> > 
> > Hmm. I thought we do *that* one ourselves. I thought we need
> > libiconv just for the other charsets.
> 
> Ohh - yes you're right - we do several builtin, by preference
> too. I thought we used to always call iconv by preference and
> then fall back if we couldn't find it, but that's not right.
> 
> We have:
> 
> "UTF8", "ASCII", "ISO-8859-1" buit in...
So, is the following correct?

- Samba 3 uses UTF8 on the filesystem side (read&write filenames) by default
for linux filsystems (ext3, ext4, reiserfs).
- On the client side Samba translates filenames from the client encoding to
UTF8 and vice versa. 
- Modern linux distributions use UTF8 encoding
- Windows XP/Vista/7 uses ... UCS2 ?
- MAC OS uses ... UTF8 ?

And thus one does not need iconv unless Samba needs to write/read file
systems that have a filename encoding that is not built-in in Samba (e.g. DOS
codepages, ISO-8859-15, etc.)

Maybe I misinterpret something here?

Thanks

On Mon, Nov 02, 2009 at 09:45:37PM +0100, theHog wrote:
> Hi,
> 
> I am building a NAS that only has a couple of SATA disks attached (no USB,
no
> CD/DVD, no NFS, no ...). Since the available space for the rootfs is
limited
> (only 8 MB), I'd like to omit libiconv because this library takes 875
Kbytes.
> 
> The filesystem on the disks will be ext3. UClibc is built with WCHAR
support.
> Can I build Samba 3.3.8 without libiconv and still have unicode filenames
> (russian, english, ...) displayed correctly in Windows and Linux clients?
No. Samba needs libiconv to do the utf8 <--> UCS2 translations for
clients.

On Tue, Nov 03, 2009 at 01:43:42PM -0800, Jeremy Allison
wrote:
> No. Samba needs libiconv to do the utf8 <--> UCS2 translations for
> clients.
Hmm. I thought we do *that* one ourselves. I thought we need
libiconv just for the other charsets.

Volker
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: not available
URL:
<http://lists.samba.org/pipermail/samba/attachments/20091103/295d47d0/attachment.pgp>

----- Oorspronkelijk bericht -----
From: Volker Lendecke <Volker.Lendecke at SerNet.DE>
To: theHog <lists at t3i.nl>
Date: 05-Nov-2009 10:29
Subject: Re: Re: [Samba] samba & libiconv
> On Tue, Nov 03, 2009 at 09:49:02PM +0100, theHog wrote:
> > So, is the following correct?
> > 
> > - Samba 3 uses UTF8 on the filesystem side (read&write filenames)
by
default
> > for linux filsystems (ext3, ext4, reiserfs).
> > - On the client side Samba translates filenames from the client
encoding
to
> > UTF8 and vice versa. 
> > - Modern linux distributions use UTF8 encoding
> > - Windows XP/Vista/7 uses ... UCS2 ?
> > - MAC OS uses ... UTF8 ?
> > 
> > And thus one does not need iconv unless Samba needs to write/read file
> > systems that have a filename encoding that is not built-in in Samba
(e.g.
DOS
> > codepages, ISO-8859-15, etc.)
> 
> Sounds right, although I'm not sure the build system allows
> this properly.
Samba 3.3.8 builds OK without libiconv. The build system gives a warning that
filename charset support is limited though.

theHog.