W3C home > Mailing lists > Public > uri@w3.org > March 2006

RE: submission of "draft-wilde-sms-uri-registration-00", "draft-wilde-sms-uri-12", and "draft-wilde-sms-service-12"

From: Aaron Irvine <Aaron.Irvine@openwave.com>
Date: Mon, 13 Mar 2006 19:03:34 -0000
Message-ID: <3F2DD0FCF1E2EA4D974C6195B482FD880179971D@bfs-exch-prd1.myopwv.com>
To: "Tim Kindberg" <timothy@hpl.hp.com>, "Erik Wilde" <net.dret@dret.net>
Cc: <uri-review@ietf.org>, <uri@w3.org>, <Bennett.Marks@nokia.com>, "Ted Hardie" <hardie@qualcomm.com>, Antti Vähä-Sipilä <antti.vaha-sipila@nokia.com>, <Basavaraj.Patil@nokia.com>, <Markus.Isomaki@nokia.com>, <alastair_angwin@UK.IBM.COM>, <Ileana.Leuca@cingular.com>, <martti.ala-rantala@nokia.com>

Hi,

Tim Kindberg wrote:
> sms-uri         =  scheme ":" sms-recipients [ ":"  sms-body ]
> [...]

An idea worth considering, but ":" sms-body would be confused with potential ":" port-number, so I'd suggest if you really want to avoid the mailto-like ?body= then:

sms-uri         =  scheme ":" sms-recipients [ "/"  sms-body ]

Although no longer query-like it could have disadvantages.




>My main motivation for keeping things short and sweet is that, as you 
>may be aware, one of the principal uses for this syntax is to put sms 
>messages into 2D barcodes for reading and sending by camera phones (see 
>www.activeprint.org). The effective capacity of such a 2D code (given 
>current camera phone optics) is about 1-200 bytes so it's nice to be 
>"lean and mean".

Most of the European languages need UCS2 to access all their characters in the MES2 subset of Unicode, that's 70*length("%hh%hh") = 420 worst case (Greek).  But for Asian languages for example, it's 70*length("%hh%hh%hh") = 630.  So if lean and mean is important, the draft should reference to IRI's which are more compact, and suggest barcodes or such like should use the IRI alternative.





http://dret.net/netdret/docs/draft-wilde-sms-uri-12.txt says:
>Implementations MAY choose to silently discard (or convert) characters in  >the sms-body that are not supported by the SMS character set they are using
>to send the SMS message.

Depending on language this can be a very bad thing, as removing a diacritical here or even dropping a letter there can change the meaning of a message.  The user would not want this done silently, so the implementations SHOULD if possible ask for confirmation.

Alternatively, or in addition, there could be a parameter ";transliteration=%c4%88:Ch,%c4%89:ch" telling the implementation how to fallback on certain characters.

Another parameter, ";lang=ISO639", would be useful for text to speech for example.  But note that ;transliteration= should take precedence over ;lang= when the implementation decides what to do with characters needing fallback.





In http://dret.net/netdret/docs/draft-wilde-sms-uri-12.txt I didn't see mention of SMS port numbers.  In JSR-120 and higher versions (how Java midlets send SMS's), there is smsurl which can be of the form sms://+358401234567:6578 (allowing port numbers would be very useful, but of course there might be security concerns).



Regards,
Aaron IRVINE
Openwave.com
Received on Tuesday, 14 March 2006 07:32:59 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Thursday, 13 January 2011 12:15:36 GMT