I was thinking that useing a viral botnet for hosting and running a wikileaks type site on part somehow using tor for it would be interesting. People could "accidentally" add themselves to the bot net. and some people could keep propaine tanks surrounding it with a small boobietrap, using Internet wifi.
or even a information leaking and hosting satalight would be interesting.
Seems like i mostly agree with 5-Justin.Wick…
90% of problems now not in hardware performance. But in modern OS -
complicated structure, preemptive multitasking (badly working “solution”
to dynamical system extensibility), unsafe-by-design multithreading (see
next sentence), bloated overdesigned libraries, layers on layers…
…note…Just as pointers, threads is oversimplified building brick, and so
not safe. (pointers+GC) is safe and is brick. Meanwhile, inheritance
also unsafe - it’s like pointer only in larger time scale, that is why
OOP is suck in large)) Because no architecture-GC in brain)) modular,
AD - was good, OOP - not.
And threads. It is also ONE-WAY dependence and so unsafe in back direction.
See? sure it has problems… many programming problems is SAME.
Uncontrolled dependencies…end-note…
“Constantly working re-simplification of all system” may be way to go))
We need implement not just “more processing power” (special purpose
coprocessors is the way for this and they already exist) but maybe some
essential language-runtime or os divide-et-conq blackmagic to rule all
this power… working on simple compiler-friendly hardware.
There are was many projects for “more high level” hardware…
Running smlnj on bare hardware seems already was done sometimes…
Lisp? is not good because 1) “only lists” is not good (memory bottleneck wise).
automatic indirection is bullshit.
2) strong static typing is a MUST, it really simplifies code reasoning.
Yes, 12-Paul, some assist to garbage collection will be nice from hardware.
This is exactly thing i want to say. not GC coprocessor with own memory
channel)) No, what would be good… is more like mlkit region-based memory
management, only segment-table alike. Think of it as of more sophisticated
MMU unit. memory protection and collection assist. sit and watch adresses.
transactional memory?
Haskell is not good (unpredictable), ML maybe good.
22-DGentry, interesting)
36-CybFreak, words like “good discipline and education could make
programmer use low level”… pfu…that just not work, and will not))
Errors is INHERENT property of ANY human brain, no matter how professional
that 1.4kg bunch of neurons is)) Real professional is just ready for
own errors… We say - “big people make big errors”, yea?))
…anyway - who controls the memory, controls the multicore world))
and remember, all there! we not need one more powerful hardware feature))
we need way to cut off this cool but unneeded brain-eating unsafe features!))
so they cant stand on our way to problem solution…
too many unneeded abstractions… no more…please…hhhh…
(silently dying)
))
s previously stated, this has always been planned -- it just wasn't a high priority relative to other things on our plate.
Regarding markdown, we have always tried to ban or convert Unicode character sets because a) spammers use them to evade our filters all the time, and b) they are annoying as hell for most users. We've also stripped out multiple spaces since the beginning of time for similar reasons.
Regarding image prefetching, I hate it because it is wasteful 99.99% of the time. That's real math and science. While it is true we use a CDN now, we pay a lot for it (protip: CloudFlare charges more than $20/month for business/enterprise customers!), and I consider it a huge waste of our resources and money. And considering I've run 4chan at a loss for a while now, it is an even bigger waste. I still ask that people not use this, as I always have, but to each their own.