12:54 p.m.
HI,
According to 23.040:
The maximum length of an uncompressed concatenated short message is 39015 (255*153)
default alphabet characters, 34170 (255*134) octets or 17085 (255*67) UCS2 characters.
But in oFono didn't match above:
For UCS2 is (255 * 66), default alphabet is (253* 153)
I file a bug to track this, let me know if it's invalid.
http://bugzilla.moblin.org/show_bug.cgi?id=8352
Thanks
Zhigang
3:11 p.m.
Hi Zhigang,
According to 23.040:
The maximum length of an uncompressed concatenated short message is 39015
(255*153) default alphabet characters, 34170 (255*134) octets or 17085
(255*67) UCS2 characters.
The structure of UD when concatentation is used is like this:
[UDH length][IEI1][IEI1LEN][IEI1DATA]PAYLOAD
One can use two header types for concatenation. 16 bit sequence and 8 bit
sequence. 16 Bit sequence uses 4 bytes while 8 bit sequence uses 3 bytes. So
the total overhead is:
1 byte for udhl + 1 byte for iei + 1 byte for length + 3|4 bytes = 7 for 16
bit header and 6 bytes for 8 bit header.
So the you provide above numbers are correct for the 8bit case, but not
correct for 16 bit case.
In 16 bit case we can encode 255 * 152 GSM characters and 255 * 66 UCS2
characters (the UCS2 characters cannot be split in the middle).
But in oFono didn't match above:
For UCS2 is (255 * 66), default alphabet is (253* 153)
oFono uses 16 bit sequence by default, so at least in the case of UCS2 it is
correct. For 7bit case you need to limit the payload to 255 * 152 characters.
253*153 will in fact result in 255 segments when using 16 bit headers.
Regards,
-Denis
8:10 p.m.
ke, 2009-12-02 kello 05:11 +0100, ext Denis Kenzior kirjoitti:
> But in oFono didn't match above:
> For UCS2 is (255 * 66), default alphabet is (253* 153)
oFono uses 16 bit sequence by default, so at least in the case of UCS2 it is
correct.
I think we should default to 8bit reference numbers, or at least make
this configurable. I've not seen 16bit ref numbers used much in the
wild.
Same applies to other "dangerous" encoding features, like language
variants. There should probably be a configuration option, or env
variable for distributors to easily enable/disable these as need be.
BTW, in smsutil.h, struct sms_assembly_node has guint8 ref, which in
sms_assembly_add_fragment() gets assigned a value from guint16.
Cheers,
Aki
3:02 a.m.
Hi Aki,
ke, 2009-12-02 kello 05:11 +0100, ext Denis Kenzior kirjoitti:
> > But in oFono didn't match above:
> > For UCS2 is (255 * 66), default alphabet is (253* 153)
>
> oFono uses 16 bit sequence by default, so at least in the case of UCS2 it
> is correct.
I think we should default to 8bit reference numbers, or at least make
this configurable. I've not seen 16bit ref numbers used much in the
wild.
3GPP 23.040 does advertise 16 bit reference headers as 'safer.' This is why I
picked it as a default. It should be trivial to use 8 bit headers if so
desired.
Same applies to other "dangerous" encoding features, like language
variants. There should probably be a configuration option, or env
variable for distributors to easily enable/disable these as need be.
We only use language variants for decoding, not encoding. However, I do agree
that we should have a global settings file for customizing some of these
options.
BTW, in smsutil.h, struct sms_assembly_node has guint8 ref, which in
sms_assembly_add_fragment() gets assigned a value from guint16.
Yes good catch. Fixed upstream now.
Regards,
-Denis
4:46 a.m.
Hi Denis,
> > > But in oFono didn't match above:
> > > For UCS2 is (255 * 66), default alphabet is (253* 153)
> >
> > oFono uses 16 bit sequence by default, so at least in the case of UCS2 it
> > is correct.
>
> I think we should default to 8bit reference numbers, or at least make
> this configurable. I've not seen 16bit ref numbers used much in the
> wild.
3GPP 23.040 does advertise 16 bit reference headers as 'safer.' This is why I
picked it as a default. It should be trivial to use 8 bit headers if so
desired.
>
> Same applies to other "dangerous" encoding features, like language
> variants. There should probably be a configuration option, or env
> variable for distributors to easily enable/disable these as need be.
We only use language variants for decoding, not encoding. However, I do agree
that we should have a global settings file for customizing some of these
options.
it was clear that we need /etc/ofono/main.conf at some point, but before
getting into it, I prefer that we collect configurable options that we
include and their preferred defaults. So feel free to propose what
should be configured by the system integrators.
Regards
Marcel
4:55 a.m.
Hi Marcel,
it was clear that we need /etc/ofono/main.conf at some point, but
before
getting into it, I prefer that we collect configurable options that we
include and their preferred defaults. So feel free to propose what
should be configured by the system integrators.
Honestly I can't think of too many right now besides what has already been
discussed:
- 16 bit vs 8 bit reference number for SMS concatenation headers
- GSM dialect for encoding (this is currently not used)
Hence this is a low priority item until this list becomes a bit more
substantial.
Regards,
-Denis
6:12 p.m.
ke, 2009-12-02 kello 18:55 +0100, ext Denis Kenzior kirjoitti:
> it was clear that we need /etc/ofono/main.conf at some point,
but before
> getting into it, I prefer that we collect configurable options that we
> include and their preferred defaults. So feel free to propose what
> should be configured by the system integrators.
Honestly I can't think of too many right now besides what has already been
discussed:
- 16 bit vs 8 bit reference number for SMS concatenation headers
- GSM dialect for encoding (this is currently not used)
Hence this is a low priority item until this list becomes a bit more
substantial.
I agree.
Besides, the first item is quite a trivial patch, I think. Attaching it
here for interested parties.
Cheers,
Aki
4672
days inactive
4673
days old
6 comments
4 participants
participants (4)
-
Aki Niemi -
Denis Kenzior -
Li, Zhigang -
Marcel Holtmann