Plans [was: Re: Gconf or GSettings? - leading to a wider question]

[Dirk Hohndel]

"Robert C. Helling" <helling at atdotde.de> writes:

> On Jan 23, 2013, at 6:06 AM, Dirk Hohndel wrote:
>
> Dirk,
>
>> Well, sort of. If you keep the mouse still and zoom out a few steps and
>> back in and out and in the position drifts. So there still is some error
>> in the math.
>
> that comes mainly from the fact that pixel coordinates are integers, so with every conversion and every division we make an error of 0.5 on the average. With each step of zooming in, everything is doubled including the error. So there is an exponential growth (like: the minimum error is the distance between two pixels in the coarsest zoom mode which is then amplified by moving in).
>
> One could try to implement our own version of the conversion functions (geo vs pixels) with floating point pixels. Then we would indeed hardwire the Mercator projection into subsurface. I don't know if it's worth it.

No, definitely not worth it. I hadn't taken the integer pixels into
account.

/D