[PATCH renderproto] Add floating point transforms

[Siarhei Siamashka]

On Fri, 15 Aug 2014 08:55:18 -0700
Keith Packard <keithp at keithp.com> wrote:

> Fixed point coordinates don't provide reasonable precision for
> transformation operations; the resulting transforms are often off by
> several pixels.
> 
> Allow clients to represent the transformation using either fixed point
> or floating point.

The protocol extension is an action with serious long lasting
consequences. XRENDER is a rather old extension and my understanding
was that it serves mostly to just allow running the existing
applications. Extending the protocol at this point needs to be
clearly justified. The current commit message is a bit vague, uses
weasel words and is not entirely accurate. So some clarifications
would be really appreciated.

* Who are the expected users of this new functionality and what are
  the practical advantages for them? Preferably something measurable
  and easy to confirm.

* Are you expecting the existing applications/toolkits to eventually
  switch from the 16.16 fixed point format to the single precision
  floating point format? Or is it only intended to be used by a few
  selected tools, such as xrandr?

* Are there any possible risks? Is there anything like like performance
  and/or accuracy drop in some use cases? Are the existing drivers and
  backends ready to provide this new functionality without exposing
  problems? Any backwards/forward compatibility concerns to care about?

* Are there any other alternatives to the proposed solution?

-- 
Best regards,
Siarhei Siamashka