NetworkManager and "illegal" SSID chars = crash?

Dan Williams dcbw at
Wed Sep 5 17:49:47 UTC 2007

On Wed, 2007-09-05 at 11:15 -0600, Richi Plana wrote:
> On Wed, 2007-09-05 at 12:59 -0400, Dan Williams wrote:
> > Right; it is an NM issue.  If NM crashes or misbehaves with a weird
> > SSID, we need to fix it.
> > 
> > As specified in the 802.11 standard, the ESSID is a 32-byte byte array,
> > there are no restrictions made as to what content that 32-bytes may
> > contain.  It is up to the program to attempt to coerce that value into
> > something to present to the user.
> Since no semantics has been proposed, then it stands to reason that one
> can just assume 1 byte = 1 character and a straight "byte value =
> Unicode value" conversion should be adopted, right? (No codepage
> conversions).
> Come to think of it, if they're just a generic 32-byte byte array, why
> even convert them to strings? They only make sense if they happen to
> contain alphanumeric and some symbolic characters (which isn't even a
> convention or restriction). Alright, maybe it's useful for displaying to
> humans when they just happen to be alphanumeric. Forget I asked.

Well, it's a really tricky problem.  I'd personally do:

try UTF-8 validation
try validation from the encoding in LANG
fall back to a lot of ? for unprintables

For example, if I live in Shanghai, I should be able to make my AP SSID
some native character like 虚.  However, if I live in Central Europoe
and for some reason I use ISO 8859-15 (or whatever it is) I'd probably
like my ň to show up.  I believe NM should try at least a few ways of
parsing the SSID (not too many) and just fall back to ? in the end.
This is what it currently does, but apparently somewhat badly.


More information about the fedora-devel-list mailing list