Re: [PATCH] GSM SMS reception code

[Dan Williams <dcbw redhat com>]

On Thu, 2011-04-14 at 15:39 -0700, Marcel Holtmann wrote:
> Hi Dan,
> 
> > 2) Testcases for PDU decoding; best way to do this is to split out the
> > bits of the SMS stuff that doesn't depend on anything MM-related like
> > MMCallbackInfo and whatnot into a separate file, like mm-sms-utils.c or
> > something, and build that file into libmodem_helpers_la_SOURCES in
> > Makefile.am.  That makes it easier to link that code to the testcases
> > without including a bunch of random stuff that isn't related to the unit
> > tests.  Then we grab a bunch of real-world encoded SMS PDUs and shove
> > them through the decoder and make sure it works on those, then we can
> > start fuzzing it too and make sure it's more robust.  We'll really want
> > testcases as there's quite a bit of variation between devices and
> > frankly nobody interprets the standards the same way :(
> 
> actually the SMS PDUs are pretty much standard since they are
> transmitted this way over the air.
> 
> You are not gonna like this, bu you guys could save yourselves some
> trouble and just use src/smsutil.c from oFono since we have been through
> this whole exercise. Including the unit tests actually.

Probably, yeah.  There's a slight question of licensing since oFono is
GPLv2 while MM is currently v2+.  Still, it already is implemented in
oFono and the license thing isn't too big of an issue.

> > 4) The whole index thing, figuring out if we can use a generic index
> > that's MM specific and maps internally to SIM indexes, or whether we can
> > in fact just use the SIM index...
> 
> You are going to enter a world of pain if you are trying to deal with
> SIM indexes or any kind of indexes for that matter. This whole indexes
> concept is pretty much ancient from around 1995. Even modern SIM cards
> are too limited for doing this properly.

Right, which is why I suggested instead using MM-specific indexes for
access via the D-Bus API and doing whatever is required internally to
present a clean external UI.  I believe the indexes in the MM D-Bus API
should not reference actual SIM messages, but should be synthetic.  I
don't think it's useful to present each SMS as a D-Bus object, and thus
indexes are the way to go.  But the indexes should be MM specific,
transient, and have no relationship to whatever happens inside MM to
operate on the SMSes.

Dan