[linux-dvb] Re: MT352 i2c register writing patch

[Holger Waechtler <holger@qanu.de>]

Patrick Boettcher wrote:

> On Wed, 5 Jan 2005, Andreas Oberritter wrote:
>
> > On Wed, 2005-01-05 at 13:35 +0100, Patrick Boettcher wrote:
> >
> > > I think to put the splitting of connected mt352-writes into the dibusb
> > > driver is the wrong place.
> >
> > Why? For me dvb-dibusb-fe-i2c.c looks like the right place because it
> > implements the i2c algorithm. It may be easier to change the mt352
> > driver, but you will have to change all other demod drivers once someone
> > decides to build a dibusb device with them.
>
> If I would implement this in the dibusb driver I need to know the size 
> of a register and its value (dib3000-frontends have 16 bit addresses 
> and 16 bit values, mt352 uses 8Bit for both). Then the questions would 
> be where to put this information? I would say this belongs to the 
> frontend-driver, wouldn't it?
>
> I don't really want to hardwire this in the dibusb driver...

can you please elaborate a little more in detail what you are planning 
to do? For engineering purposes the userspace i2c device API is more 
than sufficient, every non-defined API extension you are incorporating 
here sounds like a rape of the original idea of a consistent API for all 
devices to me. What functionality do you exactly need? If it is 
supported by more than one particular piece of hardware and interesting 
to others, then it should get a "real" public API extension, if it not 
interesting for anyone but two or three persons it should better go into 
a seperate branch but not in the public driver, not -- what do you think?

Holger