[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]

Re: NetworkManager and "illegal" SSID chars = crash?

On Fri, 2007-09-07 at 10:32 -0400, Owen Taylor wrote:
> On 9/7/07, Matthew Miller <mattdm mattdm org> wrote:
> > On Fri, Sep 07, 2007 at 08:51:30AM -0400, Colin Walters wrote:
> > > Guys, this thread is a lot more appropriate on a Bugzilla.  Say, here:
> >
> > I hate to clutter up bugzilla with vague bugs. It's not helpful for anyone.
> > I wanted to narrow things down first.
> >
> > > http://bugzilla.gnome.org/show_bug.cgi?id=474543
> >
> > <https://bugzilla.redhat.com/show_bug.cgi?id=281071>
> >
> > > (spotted at least one bug from your backtrace that's probably causing the
> > > crash; the encoding issue is a whole other set to which Owen's reply is the
> > > best)
> >
> > The crash is, however, *very* related to the encoding issue, because the
> > assumed ascii string encoding is used not only for display but as the
> > internal representation passed around in the program.
> I'd consider the two things actually quite distinct:
>  * D-BUS object path: you want a *unique* *consistent* representation of the
>    bytes of the ESSID. So you might, say, treat them as bytes and do URI-style
>    %XY escaping of bytes that aren't legal ascii characters in the context

The decision to use the SSID as part of the D-Bus object path was
insanely wrong, which is of course clear now 3 years later :)  Rectified
with 0.7, we may be able to fix it in 0.6.5 too since object paths are
essentially just pointers and are opaque.


>  * Display to the user: you want to make a best guess at the encoding and try
>    to display the exact same thing that someone typed into the router
> configuration,
>    rather some unique-but-looks-like-garbage escaped version.
> - Owen

[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]