Lucas Adamski wrote:
> I'm not sure we have any disagreement then.  Per
> https://wiki.mozilla.org/WebAPI/Security/SMS showing a user
> confirmation is a recommended mitigation, so *if* we could implement
> something like:
> explicit permission for SMS access + OS confirmation on send
> - would that work?

I expect that <a href='sms:+1234567890?body=hello%20there'>Send SMS</a> should 
open up a pre-filled SMS message, even in untrusted web content. And, I guess 
that must be what the intent mechanism does. 

Would calling send() with "OS confirmation on send" be different, UX-wise? If 
it would be the same UX, I think it would be better to simply avoid exposing 
send() to non-certified apps, to encourage developers to use the same 
mechanisms that are more likely to work cross-browser (sms:) and/or that work 
without being "privileged" and which we could potentially implement easier in a 
cross-platform way (intents). This way, send() would not operate differently 
depending on the privilege level of the app.

Also, I just now understand that there's not a separate permission for 
reading/deleting existing SMS vs. sending new SMS. While I think there are not 
many apps that want to read/delete messages but do not want to send them, I 
think there are quite a few use cases for being able to send an SMS but never 
needing to read/delete them. And, there are significant risks to allowing an 
app to read SMS (e.g. the two-factor authentication and password reset cases 
already mentioned on the wiki, and general privacy risks). So, I recommend 
separating the permission for sending from the permission from reading/deleting.

Cheers,
Brian
_______________________________________________
dev-webapps mailing list
[email protected]
https://lists.mozilla.org/listinfo/dev-webapps

Reply via email to