Re: Encoding in Kannel (was: charset question (Greek))

> -----Original Message-----
> From: Andreas Fink [mailto:[email protected]]
> Sent: Wednesday, March 13, 2002 11:23 AM
> To: Oded Arbel
> Cc: [email protected]
> Subject: RE: charset question (Greek)

> >1. Using what character set do messages arive to 
> smsbox/wapbox/mmsbox ?
> >2. Do the boxes do recoding ?
> >3. Using which character set do messages get sent to the bearerbox ?
> >4. Does the bearebox do recoding ?
> >5. Using which character set do messages arrive in the modules ?
> >6. Should the modules do recoding (except for 
> driver-specific issues) ?
> >7. How does the coding member of the Msg* structure relate to the
> >character set of the message ?
> >8. All the above questions for the MO path.
> >9. Which encodings do Kannel need to support internally ?
> >10. Using which encoding should keyword matching be done ?
> >
> >I would really like to see a clear design work on that.
> 
> 
> I agree with you in the sense that current recoding is messy.
> I would suggest that recoding doesnt occur at all unless its 
> necessary.
> This means that every SMSC driver would need to be aware of all 
> coding types and recode once from source to target if needed.
> 
> Thins like UTF16->iso->gsm should not occur as it is not "lossless".

I'd like to opt for recoding to some unicode format, as early as
possible, do everything internally with unicode, and let the modules
recode to what ever they thing is most efficient. we should supply some
functions and/or hints in the msg structure to help modules determine
which encoding would cause the least amount of data loss.
UTF-16 or UCS2 are good for that, also UTF-8. note that UTF-16 != UCS2
(!!)

--
Oded Arbel
m-Wise Inc.
[email protected]

"A witty saying proves nothing."
	-- Voltaire