The goal of libqmi?

[Dan Williams]

On Wed, 2011-10-12 at 00:23 +0200, Markus Gothe wrote:
> Afaiu the goal is to re-implement the existing oss GobiNet(3K)-exposéd QMI? Right?

Yes.  The current GobiAPI and GobiNet driver are incredibly complicated,
and we can't think of a reason why that needs to be the case.  Most
devices use some sort of TLV command structure in the firmware API, and
the Gobi API appears to hugely complicate that pattern.  In short, like
most Windows -> Linux code ports, the Gobi stuff appears to be hugely
over complicated and obfuscated, have un-necessary abstraction layers,
and reimplement common types that are already done elsewhere.

> I think documenting and find out diffrent key parameters is one thing to accomplish as well.

Definitely.  Once we get the database reader bits working better
(patches accepted!) and get a full dump of the useful bits we can start
to document the protocol.  After that we can just apply diffs from
further GobiAPI updates published by Qualcomm.

> The existing GobiNet/libqmi is exposing a "flawed" protocol (with severals parameters unknown), defined in LE (little endian) and must be patch to work with BE-systems... Just a few notes I've made in terms of improvement.

Certainly.  The code is pretty ugly.  The goal is to make the protocol
much easier to understand, much more apparent, and much easier to
program against.

Dan

> I am looking forward for your feedback.
> 
> Regards,
> //Markus - The panama-hat hacker
> 
> _______________________________________________
> libqmi-devel mailing list
> libqmi-devel at lists.freedesktop.org
> http://lists.freedesktop.org/mailman/listinfo/libqmi-devel