freenode/#sicl - IRC Chatlog
Search
12:49:30
beach
This *is* major progress, because now the entire machinery for the next phase can be loaded from files containing DEFCLASS, DEFGENERIC, DEFMETHOD, and DEFUN forms.
12:51:56
beach
frodef: I don't quite remember what you said when you described your objectives, but I do remember that they were very different from mine. Now, if you are flexible when it comes to those objectives, perhaps you find some interesting aspects in that specification.
15:41:12
frodef
I agree with many of the weaknesses of the unix-like systems of the 70's that you point out, and I think the consequences of this heritage have been rather disasterous in many ways.
15:50:26
beach
But the worst thing is probably that there are now many generations who think Unix is the best there is, and the best that there could be.
16:03:41
scymtym
i watched this last december: https://mirror.eu.oneandone.net/projects/media.ccc.de/congress/2017/h264-hd/34c3-9225-eng-How_risky_is_the_software_you_use_hd.mp4 . in a nutshell, they do automatic security assessment of binaries. the table at 15:38 shows their criteria and it is only about mitigation
16:05:23
scymtym
it is interesting, especially the (ambitious) analysis framework they propose at the end
16:06:08
scymtym
but that part struck me as strange. why would you assess the security based on the amount mitigation and damage control strategies?
16:10:51
scymtym
yeah. that is probably also the best safeguard against me misrepresenting their work (since i watched the video almost a year ago)
16:28:49
scymtym
the criteria for rating a software's security are "does it use ASLR?", "does it use stack canaries?", etc. those only mitigate the effects of the underlying actual problems
17:54:48
frodef
My historical understanding of unix (multics) processes was perhaps more that it was a way to share computers between users, i.e. each process simulating basically an isolated, separate computer.
17:56:00
frodef
(The more recent VM emphasis being basically the same pattern repeated with yet another level of indirection.)
19:04:17
jcowan
The earliest Unix had no concept of users, being in fact a single-user machine. The point was to provide a functional meta-language that represented stateless computation between arbitrary black boxes.
19:05:04
jcowan
The black boxes could compute with effects and generally did, but no effect could spread beyond a certain scope, short of kill().
19:06:28
jcowan
Multics did indeed have one process per user as a rule, which is where the shell got its name; it was literally a shell that dynamically linked the code you wanted to run into itself, all programs being represented by libraries with a fixed interface.