NetworkManager and "illegal" SSID chars = crash?

Dan Williams dcbw at redhat.com
Wed Sep 5 19:06:06 UTC 2007


On Wed, 2007-09-05 at 14:57 -0400, Owen Taylor wrote:
> On 9/5/07, Richi Plana <myfedora at richip.dhs.org> wrote:
> > On Wed, 2007-09-05 at 12:44 -0400, Owen Taylor wrote:
> > > > Shouldn't those octet character strings be converted to Unicode strings,
> > > > anyway? I thought that Gtk (via Pango) had displaying international
> > > > character down pat.
> > >
> > > In order to "convert octet character strings" into Unicode and display
> > > them, you have to know what encoding the strings are in. Bytes are
> > > just bytes until you have an encoding.
> >
> > I was under the assumption they were 8-bit ASCII strings.
> 
> There's actually no 8-bit ASCII. People sometimes use that to mean
> Windows codepage 1252
> (http://en.wikipedia.org/wiki/Windows_code_page).
> 
> > Anyway, it's a NetworkManager issue. They would know what the drivers are returning,
> > right?
> 
> The drivers may well just be returning whatever bytes were sent over
> the wire. Now, in any *sane* protocol, the protocol docs would define
> what encoding the bytes over the wire are in, but it took a very long
> time for people to realize this, so it wouldn't shock me if 802.11b
> doesn't specify. Then you'd basically have to do experiments with
> common access points to see how they encode things, then add
> appropriate heuristics to NetworkManager.

The drivers return a 32-byte array containing arbitrary values to
NetworkManager, which is technically correct.  It's up to NM and
userspace to deal with this somehow.  802.11 specifies nothing about
encoding or anything other than a max length of 32 bytes.

Dan





More information about the fedora-devel-list mailing list