IRC channel logs

2021-05-06.log

back to list of logs

<fossy>afternoon slang in australia
<fossy>I habitually type it and then I remember most ppl arent Australian
<OriansJ>fossy: well most people don't write in proper english either. So as long as everyone is willing to clarify to aid in understanding, I'm sure it'll be fine.
<xentrac>oh interesting, I had no idea. there's a crazy antidemocratic network protocol called arvo too
<OriansJ>xentrac: there are even rule34 entries for avro as well but that honestly isn't even surprising at this point.
<melg8>Thanks, @fossy i've been watching SN15 test.
<melg8>@fossy you can link it with my nick on message - i'll check out logs after i'm back.
***edef_ is now known as edef
<bauen1>annoying, tar tries to use the tar archies uid/gid when extracting as root, since at least version 1.34
<bauen1>and that breaks inside user namespaces for obvious reasons
<bauen1>the option to disable this behavior is however not recognized by the earlier tar we're using ...
<Hagfish>oof. does it need a chown afterwards?
<bauen1>well, if you're actually running live-bootstrap as root it's not a problem (except having some uid 1006 files laying around after unpacking gmp)
<bauen1>but if the chown fails tar will exit with != 0, breaking my user namespace build after we've build tar 1.34
<melg8>Hi, i've was thinking yesterday, it it possible, and is it even viable approach - to combine something like stage0 bootstrapping with homomorphic encryption, splitting data provider and data executor,so host system executing computation even if it has some kind of compromise  behavior would not be able to detect that it should subvert results of
<melg8>computation?
<bauen1>melg8: you should take a look at the logs, i've already made a few brain dumps into this channel about how you can improve the security of the bootstrap
<melg8>thanks, i will!
<stikonas>although, you have to trust your host at some point
<stikonas>it's the hardware on which you run things
<bauen1>i'd have to take another look at how cross compiling can be used to prevent trusting trust attacks
<stikonas>if it's backdoored, you are in big trouble even if the rest of bootstrap is trustworthy
<bauen1>you can "outsource" many tasks, but you need some form of computer that is trusted at least (doesn't have to be fast)
<bauen1>it's kind of like SAT solving (or something like that) where the programs used to compute a proof can be untrusted and you only need to trust a very small very simple and verified program that verifies the proof
<melg8>another thing i've been thinking -is that if system is turing complete - it can implement some form of bootstrap in it. So like - game of live running stage0.
<gforce_de1977> http://ironcurtain.svelch.com/wp-content/uploads/sites/4/2018/12/obr041-1024x703.jpg
<gforce_de1977>she knows, how to bootstrap
<stikonas>well, in the early days of computing there was no gender misbalance, so probably it was quite usual sight
<vagrantc>that's a bold claim
<vagrantc>lack of proper credit definitely made it hidden over time, but *no* gender misbalance?
<stikonas>vagrantc: wasn't gender misbalance caused by universities offering IT courses?
<stikonas>I thought before when programmers were pulled from other professions, it was much more balanced
<vagrantc>more balanced, i would guage
<stikonas>well, probably actually more woman were programmers back then
<stikonas>so maybe misbalanced in the other direction
<melg8>Here we go https://github.com/fosslinux/live-bootstrap/pull/115 - something back to community from me.
<vagrantc> https://debconf18.debconf.org/talks/108-in-1968-mom-built-a-computer/