20 Jan 2021 |
Artifth | Also, dotnet had something similar, where VM could load different programs into same process and provide isolation by itself | 13:38:50 |
Artifth | They killed it in Core | 13:38:58 |
Artifth | In reply to @artifth:feneas.org Also, dotnet had something similar, where VM could load different programs into same process and provide isolation by itself It also could isolate assembly from doing any system access, thus allowing safe execution of untrusted code | 13:42:23 |
Artifth | https://stackoverflow.com/questions/1094478/what-is-a-net-application-domain | 13:43:50 |
Artifth | Wonder how computing world would look had MS not failed implementing original COM+/.NET vision | 13:44:43 |
Levy (Elara) | In reply to @artifth:feneas.org Good luck fighting VM escape bugs For embedded it's not as bad as long as you avoid RCE bugs | 13:48:03 |
Artifth | I'm thinking general computing | 13:49:39 |
Artifth | Like Android, but with kernel-level Dalvik | 13:50:00 |
Levy (Elara) | yeah... there it might not be as nice | 13:50:28 |
| * Levy (Elara) finds it amusing that MS has at one point tried all of the nice things but failed to implement them properly ending in someone else picking it up 10-20 years later | 18:09:37 |
informer | https://www.jsof-tech.com/disclosures/dnspooq/ | 18:33:18 |
informer | hmm is there any way to check if my router uses dnsmasq? | 18:54:23 |
Artifth | In reply to @tauoverpi:feneas.org finds it amusing that MS has at one point tried all of the nice things but failed to implement them properly ending in someone else picking it up 10-20 years later Welp, judging by their financial reports, half-assing through nice ideas seems to work well | 20:28:04 |
Artifth | In reply to @tauoverpi:feneas.org this also allows for multi-tasking while compiling in points which call out to the VM every now and again (yes, it's effectively syscalls but without the expensive context switches) Could we, in theory, implement such VM using native x86 instruction set as a byte code? We'll need some kind of x86 to x86 JIT that would rewrite all memory access instructions | 22:57:05 |
Levy (Elara) | You could though it'd be rather difficult | 22:59:44 |
Levy (Elara) | not with PIE though | 23:00:04 |
Levy (Elara) | * You could though it'd be rather difficult | 23:01:15 |
Levy (Elara) | * not with PIE though | 23:01:24 |
Levy (Elara) | This isn't that difficult now that I think of it | 23:01:40 |
Levy (Elara) | the main issue is the size of it | 23:01:48 |
Levy (Elara) | you'd also need to emit bounds checks | 23:03:01 |
Artifth | In reply to @tauoverpi:feneas.org For embedded it's not as bad as long as you avoid RCE bugs Also VMS for embedded that kinda do this already exists. Stuff like micropython or dotnet nanoframework | 23:10:41 |
Levy (Elara) | yes but they're language bound | 23:14:36 |
Levy (Elara) | and the languages tend to suck | 23:14:41 |
Levy (Elara) | https://metacall.io/ | 23:16:12 |
Artifth | In reply to @tauoverpi:feneas.org yes but they're language bound Well, need to make one for WASM. It seems to be targeted by many languages nowadays | 23:20:00 |
Levy (Elara) | There are a few attempts | 23:20:30 |
Levy (Elara) | there's also eWasm which targets embedded | 23:20:40 |
Levy (Elara) | though (afaik) it's just for one application | 23:21:16 |
Levy (Elara) | In reply to @artifth:feneas.org Well, need to make one for WASM. It seems to be targeted by many languages nowadays with support for WASI or some form of it | 23:21:37 |