Conversation

Notices

1. liate 💾 (liate@hackers.town)'s status on Sunday, 09-Jul-2017 19:34:06 UTC liate 💾

   • Hallå Kitteh

   Cool article (from @clacke )
   Warning: clickbait title
   Privileged Ports are Causing Climate Change
   http://adamierymenko.com/privileged-ports-are-causing-climate-change/

   The basic idea is that privileged ports have led to such hacks as virtualization and containerization for multiuser use for hosting and stuff, and a system which actually had multiuser capability for networking (as in, let everyone use all ports) would be much more resource-efficient.

   Thoughts on the article?

   • Hallå Kitteh repeated this.
   • Hallå Kitteh (clacke@social.heldscal.la)'s status on Monday, 10-Jul-2017 04:18:55 UTC Hallå Kitteh
     @liate The canonical term is "/via". :-)
   • Hallå Kitteh (clacke@social.heldscal.la)'s status on Monday, 10-Jul-2017 04:26:33 UTC Hallå Kitteh

     • lhark
     @lhark @liate On the other hand, needing root privileges just to put a server on port < 1024 opens up for privilege escalation vulnerabilities. In theory it doesn't, because bind, fork, deprivilege should be a solved problem, but in practice it is a problem.
     
     A system with more fine-grained capabilities and better APIs would help a lot against all kinds of things, and in that context, limiting ports < 1024 to root is pretty arbitrary. Either allow any port, or require privileges per port: User (or process) X is allowed to listen on interface Y port Z.
   • Hallå Kitteh (clacke@social.heldscal.la)'s status on Monday, 10-Jul-2017 04:28:16 UTC Hallå Kitteh

     • lhark
     @lhark @liate
     
     > But hypervisor escape is several magnitudes harder.
     
     I don't know. Either there is an escape or there isn't. Once an exploit is there, it's trivial to use. A hypervisor is a really complex piece of software, so I don't know that you can say that it has a smaller potential for vulnerabilities.
   • Hallå Kitteh (clacke@social.heldscal.la)'s status on Monday, 10-Jul-2017 05:09:59 UTC Hallå Kitteh

     • lhark
     @lhark @liate
     
     Yeah, that's why I mentioned evolutionary path dependence in https://social.heldscal.la/notice/2937355 . It's really hard to get out of a local optimum*.
     
     * Not literally an optimum. Maybe a Pareto optimum if you count all the developers that would have to rework their software as you start breaking APIs to improve things.
   • Hallå Kitteh (clacke@social.heldscal.la)'s status on Monday, 10-Jul-2017 05:15:20 UTC Hallå Kitteh

     • lhark
     @liate @lhark I guess what can be said about hypervisors is that maybe it's less possible to take shortcuts that threaten security, because the interface is small and generic.
     
     But then we have virtio drivers, and I don't know enough about this but it seems to me that as we try to improve performance, we introduce larger attack surfaces, thereby closing the gap between virtualization and containerization in more than one way.
     
     From what I understand, people like Joyent are closing the gap from the other direction in an interesting way, running Linux apps in Solaris zones, which are older and cleaner than Linux namespaces. They're confident enough in it that they are doing this while co-hosting customers.