On 04/06/2015 23:59, Alex J Lennon wrote:

On 04/06/2015 23:03, Denis Kenzior wrote:
Hi Alex,

Ordering should have nothing to do with it.

Yes, the ordering is relevant. We (like other ofono users I suspect)
have to allow multiple APNs or the automatic provisioning process fails.

Then, the first context found in serviceproviders.xml is what is used by
default for the connection.

An example of the problem is that if you use a major telco's SIM card in
the UK - Vodafone, ofono will then default to using an ASDA mobile
context because of the ordering, and this will fail.

My feeling is that a larger provider like Vodafone or O2 should be the
default, not ASDA mobile or GiffGaff, and this should thus come first
(understanding that the Ofono project does not control this document)
It has been years since I wrote the provisioning plugin, but the
intent was to fail if looking up MCC/MNC combo resulted in multiple
matches. So this may be a bug, or you might be using some custom
behavior.  But in the end, ordering of the entries should not affect
the provisioning logic.

Allowing Duplicates - Not by default no, but you have a boolean
parameter in there and logic to allow for duplicate contexts, which we
have to enable (as do others I think from my Googling on this) or the
provisioning support is unusable with the upstream serviceproviders.xml
as far as I can see.
Then that's the problem.  The intent was never to allow duplicates.
That boolean was added for tools/lookup-apn only.

I'm not entirely sure how the RilModem fork relates to Ofono but you can
see they had the same problem


    * TODO: review with upstream. Default behavior was to

    * disallow duplicate APN entries, which unfortunately exist

    * in the mobile-broadband-provider-info db.



SPN - Thanks. This seems promising. I will investigate the SPN values
The real fix is to fix mobile-broadband-provider-info.
Yes I would agree with that.

As I come to investigate this, I find I am concerned about using the
Service Provider Name as I can't see any registry for those names, it's
free text for display purposes, so I assume it is at least possible it
might change without warning,
whereas there does seem to be a registry for MCC/MNC (e.g.

I am thinking it may be preferable to use the registered IIN number from
the ICCID - http://www.controlf.net/iccid/

This seems a more controlled way of providing the uniqueness needed to
me and presumably it's easy enough to read the ICCID out, if it's not
already being read out.

No that's not going to work. I see ICCID prefix is the same for e.g. O2 and Tesco Mobile, being MCC and MNC






Tesco Mobile
(MVNO of O2)