|
im a decently smart person and the last pages of posts about webass have caused me to be more confused about it rather than less
|
#
?
Dec 28, 2017 01:31
|
|
|
|
| # ? May 15, 2024 21:00 |
|
|
This sounds a mess, MPI based Numpy in Google  quote:STAC recently tested two clusters using STAC-A3 benchmarks: the STAC-A3 Pack for Python (Rev A) using Python 3.6.2 and the High Performance Analytics Toolkit (HPAT) on Google Cloud 64 vCPU Intel Xeon (Skylake) instances with 14TB of Google Persistent SSD. One cluster (HPAT171028) contained 5 compute nodes, while the other (HPAT171029) contained 20.
|
|
#
?
Dec 28, 2017 02:14
|
|
|
Thermopyle posted:im a decently smart person and the last pages of posts about webass have caused me to be more confused about it rather than less there was very little talk about webass in the past few pages, mostly binast talk, so iunno how that happened. what are you confused about
|
|
#
?
Dec 28, 2017 02:21
|
|
|
if thats webass whats webtp
|
|
#
?
Dec 28, 2017 03:09
|
|
|
creatine posted:if thats webass
|
|
#
?
Dec 28, 2017 04:27
|
|
|
Peeny Cheez posted:A Javascript interface to popular excretional database MyBNGHL lmao 5
|
|
#
?
Dec 28, 2017 08:14
|
|
|
Soricidus posted:no you see there are very good reasons why I need to implement my own email address validation instead of just having it built into the browser the browser might implement email address validation correctly and it is very important that I use my bespoke solution that rejects valid addresses
|
|
#
?
Dec 28, 2017 08:37
|
|
|
Peeny Cheez posted:A Javascript interface to popular excretional database MyBNGHL
|
|
#
?
Dec 28, 2017 18:17
|
|
|
and some sauce-a on my balls
|
|
#
?
Dec 28, 2017 18:29
|
|
|
what I really like about the Binary AST proposal is it’s putting safety directly at odds with performance, as performance is its sole reason for existence. and does it have a high-level strongly-typed set of APIs you might be able to secure? no, it takes un untyped blob of attacker-controlled bytes, performs as little validation as possible and puts it straight in the middle of code generation and execution  now that’s irresponsible software engineering at its finest  the real reason WebAssembly is useless is that with Binary AST and one wrongly generated jump instruction you can just execute real assembly instead
|
|
#
?
Dec 28, 2017 19:24
|
|
|
Max Facetime posted:what I really like about the Binary AST proposal is it’s putting safety directly at odds with performance, as performance is its sole reason for existence. and does it have a high-level strongly-typed set of APIs you might be able to secure? no, it takes un untyped blob of attacker-controlled bytes, performs as little validation as possible and puts it straight in the middle of code generation and execution binary ast comes before code generation
|
|
#
?
Dec 28, 2017 19:45
|
|
|
like, you know what an AST is, right? there's nothing malicious an attacker can do by manipulating an AST that they can't do through the syntax of the language (at least in JS).
|
|
#
?
Dec 28, 2017 19:46
|
|
|
i'm the security checks done during parsing wait, this is js, i might actually exist
|
|
#
?
Dec 29, 2017 00:18
|
|
|
can't wait for webass bitcoin miners
|
|
#
?
Dec 29, 2017 00:19
|
|
|
raminasi posted:i'm the security checks done during parsing reminder that windows defender has a process running with root privileges that evals every single piece of javascript that enters your computer just in case it might do something malicious
|
|
#
?
Dec 29, 2017 02:33
|
|
|
my graphics card driver runs not one, not two, but four instances of nodejs... i assume for massively parallel workloads?
|
|
#
?
Dec 29, 2017 07:04
|
|
|
Max Facetime posted:what I really like about the Binary AST proposal is it’s putting safety directly at odds with performance, as performance is its sole reason for existence. and does it have a high-level strongly-typed set of APIs you might be able to secure? no, it takes un untyped blob of attacker-controlled bytes, performs as little validation as possible and puts it straight in the middle of code generation and execution sandboxing some code that does arithmetic on raw memory bytes is way the hell easier than sandboxing almost anything else it's an increase in the attack surface, but so is everything with the web. pound-for-pound, it's very secure you can get the same one-jump-wrong stuff with normal javascript too
|
|
#
?
Dec 29, 2017 08:34
|
|
|
yeah after this discussion i have even less of an idea what webass is supposed to do or be
|
|
#
?
Dec 29, 2017 15:57
|
|
|
Themage posted:my graphics card driver runs not one, not two, but four instances of nodejs... what graphics card driver does that?
|
|
#
?
Dec 29, 2017 16:16
|
|
|
Themage posted:my graphics card driver runs not one, not two, but four instances of nodejs...
|
|
#
?
Dec 29, 2017 16:19
|
|
|
I've heard of bloated printer support programs running apache and poo poo, but not graphics card drivers. Maybe that lovely nvidia experience app is electron?
|
|
#
?
Dec 29, 2017 16:21
|
|
|
mystes posted:You mean your graphics card accelerates node.js for some reason or it comes with bloatware using electron or something? fwiw nvidia experience is node.js
|
|
#
?
Dec 29, 2017 16:22
|
|
|
cinci zoo sniper posted:fwiw nvidia experience is node.js Which of course is not the same thing as 'the driver' 
|
|
#
?
Dec 29, 2017 17:33
|
|
|
feedmegin posted:Which of course is not the same thing as 'the driver' that it isnt, henc fwiw to bloatware question. nothing forces one to install gfe
|
|
#
?
Dec 29, 2017 17:34
|
|
|
feedmegin posted:Which of course is not the same thing as 'the driver' something something close to the metal
|
|
#
?
Dec 29, 2017 18:57
|
|
|
HoboMan posted:yeah after this discussion i have even less of an idea what webass is supposed to do or be web assembly is a gaggle of nerds arguing about how to turn chicken poo poo into chicken salad
|
|
#
?
Dec 29, 2017 19:34
|
|
|
AWWNAW posted:web assembly is a gaggle of nerds arguing about how to turn chicken poo poo into chicken salad Use it to fertilize plants, then feed those plants to a chicken.
|
|
#
?
Dec 29, 2017 21:30
|
|
|
All I know is that webass doesn't make writing dom manipulations in python easy, so it's essentially worthless to me. Your options are ship the entire 3mb standard library compiled to webass, which is stupid because you barely use any of it, or to do what frontend devs have been doing for a generation which is precompile using nodejs and ship only the parts you use and then you're still using npm/grunt/yarn/regretting your life choices.
|
|
#
?
Dec 30, 2017 05:02
|
|
|
TimWinter posted:All I know is that webass doesn't make writing dom manipulations in python easy, so it's essentially worthless to me.
|
|
#
?
Dec 30, 2017 06:30
|
|
|
HoboMan posted:yeah after this discussion i have even less of an idea what webass is supposed to do or be the purpose is to make code in the browser run faster in two main ways: faster parsing, more straightforward and predictable conversion to native code but right now it has no ability to touch the dom at all, it can only compute that's about it. just ignore everyone being pedantic about ASTs
|
|
#
?
Dec 30, 2017 09:00
|
|
|
oh cool so it can't do the only thing anyone uses javascript for? good replacement
|
|
#
?
Dec 30, 2017 09:42
|
|
|
don’t worry we’ll fix it by just using webass as a platform for reimplementing everything else after all it’s pretty dumb that browsers stay as old fashioned desktop apps when everything else is a web app. I want to run firefox in chrome dammit
|
|
#
?
Dec 30, 2017 09:50
|
|
|
Soricidus posted:don’t worry we’ll fix it by just using webass as a platform for reimplementing everything else *sighs drawing katana* time to sever youre content delivery network
|
|
#
?
Dec 30, 2017 10:35
|
|
|
*watches demo of AAA game engine running in browser* ok but can I use this to place ads in the sidebar
|
|
#
?
Dec 30, 2017 15:37
|
|
|
mystes posted:How does webass inherently preclude shipping a subset of the standard library? Maybe lovely dynamic languages like Python make it too difficult to do tree shaking to remove unused code. On the other hand, tree shaking works well for non-lovely dynamic languages like Lisp, so who knows.
|
|
#
?
Dec 30, 2017 16:29
|
|
|
all tree-shaking solutions assume you’re not doing certain kinds of reflective things. like you can’t drop a type if an api lets you look it up by name. that sort of thing is uncommon in lisp, although i believe the apis still exist. i don’t know how common it is in python
|
|
#
?
Dec 30, 2017 17:04
|
|
|
Ugh, wasn't tree shaking still called DCE like a month ago?
|
|
#
?
Dec 30, 2017 18:43
|
|
|
i actually call it gdce, iirc tree shaking is just an algorithm. lispers really like conflating concepts with implementations because lispers are the wolframs of plt. anyway don’t blame me i was following the thread ofc maybe i shouldn’t admit to following a thread that just spent like five pages trying to figure out what an ast is
|
|
#
?
Dec 30, 2017 21:58
|
|
|
Dylan16807 posted:but right now it has no ability to touch the dom at all, it can only compute asm.js at least included a foreign function interface that, I assume, permitted you to call arbitrary JavaScript methods you register with parameters constrained to 32-bit integers and doubles does webass not have an ffi?
|
|
#
?
Dec 31, 2017 02:23
|
|
|
|
| # ? May 15, 2024 21:00 |
|
|
is reading my article really too hard for you guys https://gist.github.com/magcius/8959ca1f5bbde4bd66b98b6d5dc73112
|
|
#
?
Dec 31, 2017 02:29
|
|