Re: [dev] Surf update

From: <sylvain.bertrand_AT_gmail.com>
Date: Sat, 27 May 2017 12:34:14 +0000

Well,

It seems that EU (I'm french) is not able to secure a *proper* "comfort zone"
to let us release our coding power.

I did investigate EU funding seriously though, but what I got my
hands on, are jokes, real f*cking jokes.

I perceived like some "forces" are blocking any attempts at funding significant
and realistic setups for those projects (in the huge economy which is EU).

IRL matters forbid me to have a clear mind and more, that in order to code
something significant (my IRL is indeed accelerating towards total
destruction). I feel totally castrated and unable to express my skill sets: 2
people which were closed to me, literally told me that I should drop open
source software if "I don't want to die"...

That said, since the coding solution seems way to far away, and our amazing
legal system is supposed to protect the small against the big, on the WEB
front, I'm actually preparing to go the legal way on critical web sites (I'll
ask first nicely, coze not all of them are gangsters or sick minds, they just
have a lack of perspective).

Most of the critical web sites can have a noscript portal for "dirt cheap".
Those web sites with a "service" too far away from a noscript web portal (a
"rich interface is kind of mandatory"), they could provide a simple web
service/network protocol interface (which do not depend on current web engines
like this insult to sanity which is OAuth).

On the compiler front, many uses of GNU extensions may be fixed in the original
code with alternatives, but the one which is mandatory and big for performant
kernel/libc programming is in-line assembly. Replacing in-line assembly with
assembly function calls will very probably kill fast-path performance in many
cases.

On GPU shaders, no salvation, a SPIR-V compiler has to be written from scratch
and some mistakes in the specifications made it far from trivial. But it's far
from the abyssal work needed on the web front (to keep complexity in check, you
must pre-build the vector control paths from blocks).

On unicode, well, to keep it reasonable, just a lean implementation limited to
roman like languages will do for the time being... but its trash coders are
changing its interface more often than linux kernel interfaces for no reason.
(need to have a look at harfbuzz-ng though).

-- 
Sylvain
Received on Sat May 27 2017 - 14:34:14 CEST

This archive was generated by hypermail 2.3.0 : Sat May 27 2017 - 14:36:15 CEST