Re: A Critique: Getting Started with GNUstep on Windows

[Riccardo Mottola]

Hi,

Gregory Casamento wrote:
> Riccardo,
>
> On Wednesday, February 17, 2016, Riccardo Mottola 
> <riccardo.mottola@libero.it <mailto:riccardo.mottola@libero.it>> wrote:
>
>     Hi,
>
>     Richard Frith-Macdonald wrote:
>
>         It would be nice to target the Rasbpberry Pi using Wayland
>
>     me too. except for Wayland. I want to have run on raspbian
>     smoothly instead of the provided WM they have.
>
>
> Why not wayland?  Network transparency is not something which is super 
> important these days even on the rpi.   Wayland brings us a number of 
> advantages:

Right for such a small box network transparency is a good thing.
Some people may want to use a "mini computer" like the piOS guys with 
the mini NeXT CUBE, but others may want to deploy it headless.

Also, currently X11 works fine on the Pi and the raspbian is , thus my 
statement was intented for the short-term future.

I think we can support wayland along X!!? actually we absolutely should. 
If Wayland becomes interesting on the PI we can just the other backend.

What's the best way to be interesting on the PI and similar project? 
exploit their current stuff and make the best experience right now and 
gradually improve it? Or imagine a disrupitve change.. and be ready in a 
year with a new backend (which means we will fix and improve other 
things less)?

> * native compositing
> * it expects frameworks to draw their own Windows something we do 
> extremely well.
> * it supports true window transparency
> * it allows animations and such to be done much more l easily if and 
> when needed
> * etc
>
> I think we really need to question whether X truly serves our needs.  
> No one is saying leave it behind.  What we are saying is to take 
> advantage of the things wayland does for us.

Sure.. but is Wayland to help us on these small devices, to be more 
appealing there? Taking in consideration that the GPU and FPU of these 
machines aren't that fast, so effects and other gizmo need to be used 
sparingly.
My thought was not general, not on a super-workstation with 8 cores and 
twin GPUs and more videoram than the PI itself has.

> These need to be looked at.  Slowness has been a problem.  I can 
> remember using GNUstep on a 128mb Linux machine 10-12 years ago. I 
> would like to get some benchmarks which are unbiased about our 
> performance rather than hearsay and conjecture.  I would also like us 
> to focus on hardware on which GNUstep will be realistically deployed 
> rather than taking the netbsd approach and worrying about making it 
> work everywhere.

I think if would be interesting to have some sort of "gradual 
degradation" of features vs. speed. A bit like WIndows does with Aero 
depending on your hardware, but not applied only to graphics.
Of course, general optimiazions benefit everybody, no doubt.
I suppose though that people wanting to use GNUstep on their workstation 
or laptop to develop natively, would like all the features.

While the Raspberry has a decent CPU, its FPU and GPU are slow compared 
to it as sperhaps are other component. That is, if I take an older 
computer with x86 or ppc with comparable cpu, I see that it feels a lot 
snappier.
Cairo, for example, is heavy on the FPU and generally, GNUstep is quite 
heavy on it. That's the reason why the letux400 with a 360Mhz effective 
clockspeed, but no FPU is very very slow with GNUstep, slower than a 
100Mhz SPARC cpu! Other applications, which do not use cairo/art/gtk... 
can be reasonably fast.

In Dublin with Richard we found that memory usage is quite high, but 
found no obvious reason for it, not a particular dependency to disable.


Riccardo