Re: [Pan-users] How to use Numeric Pad 'Delete' key

[Rhialto]

On Thu 19 May 2016 at 18:30:28 -0700, walt wrote:
> I hope you're not as grumpy as you sound in print ;)

Hehe, not I'm quite cheerful usually :-) It's just that if one is being
honest, nothing new has been invented in computers since about 1970.  I
like to point that out when I'm playing advocate of the devil.
Everything so-called new since then is mostly hype. Internet? The
ARPAnet is from 1969. Virtual Machines? IBM did it in 1964 (IBM_CP-40).

> How do you feel about 'mir' from Canonical?

Not really a fan of that either :-) I don't see anything fundamentally
wrong with the X protocol. It's client/server with a very useful network
connection capability. Other more recent graphics systems, such as BeOS
which was built from the ground up was also client/server, so that shows
that the idea is sound. And as long as that is the case, why change the
protocol? If something is missing, create an extension. If some
extension is no longer needed, drop it. The X protocol was made
extensible with a reason.

Much of the X software is quite ahead of its time. Take the Xt
X Toolkit Intrinsics library. It defines widgets in an object-oriented
way, very much like java does, or Qt. Unfortunately, C at the time was
not object-oriented, so you have to write lots of tedious stuff by hand,
that real object oriented languages would handle for you automatically.
But there is nothing preventing anyone from writig a little tool to turn
a more user-friendly description into all the boilerplate C code. (In
fact I have been working on such a thing for a bit:
https://github.com/Rhialto/twiXt ). Recently, in the course of debugging
memory leaks in Pan, I have been diving into how GTK defines its
widgets, and actually, it isn't really much better than how Xt does it!
Only a very little bit.

Font rendering is one thing that for some reason has been mishandled by
libraries like GTK. There is a TrueType font extension, so there is
hardly any reason to render fonts client-side. Of course, when you
render stuff client-side it's slow. The server is supposed to have all
the wizz-bang accelleration hardware, so use that capability by sending
sufficiently high-level requests to it! And if the Wayland fans now say
"that's exactly what we are doing", then probably they could also have
done it with X. Run the whole X server on the graphics card, if you
want. I see no reason why you can't (after you add some appropriate
communications link for it). After all there are graphics chips that run
complete independent operating systems (the Raspberry Pi is apparently a
video chip that has a small computer attached in a side-corner
somewhere; the VideoCore boots and runs its OS first, then it starts the
ARM).

I agree that some parts of X have been over-complicated in recent years.
But I would argue that that is not because X is inherently not good
enough, but because people have been tinkering with it who don't
properly understand it, and how to do it right. (This whole Kernel Mode
Switching is a disaster: now you need drivers for your graphics card in
your kernel as well as in the X server. Idiotic.)

The X Window System is the successor of the W Window System. So one
would think that the next successor would be the Y Window System. Going
back to W sounds like going backwards :-)

Sorry for the rant :-) I'm using my oneliner from the earlier post as my
signature now, let's see what effect that will have :-)

-Olaf.
-- 
___ Olaf 'Rhialto' Seibert  -- Wayland: Those who don't understand X
\X/ rhialto/at/xs4all.nl    -- are condemned to reinvent it. Poorly.

signature.asc
Description: PGP signature