[PATCH 5/8] dt-bindings: memory: Add Tegra264 definitions

[Rob Herring]

On Thu, May 08, 2025 at 11:09:12AM +0200, Thierry Reding wrote:
> On Thu, May 08, 2025 at 10:45:29AM +0200, Krzysztof Kozlowski wrote:
> > On 08/05/2025 10:02, Thierry Reding wrote:
> > >>
> > >>
> > >>> how much more you'd like me to make it based on that. Do you expect me
> > >>> to add the nvidia, prefix? In that case it would be inconsistent with
> > >>> all of the 8 other Tegra MC includes that we have in that directory.
> > >>
> > >>
> > >> Same story as for every other case, why this would be different? All of
> > >> them - amlogic, mediatek, samsung, qcom, every soc - move to new style
> > >> since some years, why this one should be different?
> > > 
> > > Because we've used exactly this naming convention for more than a
> > > decade. I get that it's nice to have consistency, but do you really want
> > > me to go and churn all of these files just so we can add a vendor-prefix
> > > and drop a redundant suffix?
> > No, I want new files. Look:
> > 1. Some time ago tegra-1.h was added.
> > 2. Someone spotted that there was tegra-1.h, so added now tegra-2.h.
> > 3. Now this is a pattern so of course next person, even if asked to use
> > vendor prefix, will not, right? Because it would break the pattern. So
> > we have tegra-3.h
> > 4. tegra.4 - no vendor prefix, because you have already three cases.
> > 5. You see where I am going?
> > 
> > All of above - amlogic, mediatek, samsung, qcom - had decade of such
> > convention. I asked to changed, they used the same argument as you
> > ("decade") and then changed.
> > 
> > Why this is different case than decade in amlogic, mediatek, samsung and
> > qcom?
> 
> It's a matter of principle. One convention is as good as another. There
> are no clear advantages of one over another. It's pointless and frankly
> there are more important things than changing filenames that everybody
> has been okay with for the last 10 years.

The issue is one convention is consistent for you and only you, the 
other is consistent for *everyone*. And then we'll have someone argue 
they are just following the same convention as Tegra...

If you had several drivers and add a new one, would you argue that the 
new one should follow the conventions of the old drivers rather than 
current conventions? No, you wouldn't.

Rob