🚀 go-pugleaf

dbowie446@aol.com (DBowie446) wrote:
]
] >  And only marginally improved UI's are undesirable for most people
] >  at the expense of being able to run their software reliably.  So
] >  if you have to choose, anything that works is better than
] >  anything that doesn't (my favorite restatement of Occam's Razor).
] 
] The assumption there being that the vast majority of users will
] greatly benefit from Protected Memory schemes when they are
] introduced. They won't: *programmers* will. To the end user, a
] radical change in the way your machine multitasks your
] applications or juggles your available RAM will be invisible. To
] the end user, such changes are so close to completely invisible
] that they hardly matter.

The assumption here being that the vast majority of users run
bug free code and so can safely cooperatively task in an unprotected
environment without having system crashes or data loss.

An assumption which is provably false.

] Further, Occam's Razor aside, assuming
] that *any* protected memory scheme will mean the end of viruses
] or systemwide software crashes is in clear violation of my other
] favorite law, Murphy's...

I don't assume that; I've written a UNIX "virus" as part of an
advanced computing project.  You probably have code similar to
the "work" part of the "virus" on your machine now: any self
extracting archive qualifies.  The point was to compress binaries
but leave them usable.  A loopback/union mount of a file system
and some glue code makes this type of thing unnecessary, but my
code was pre-file system switch/vfsswitch (hence the choice of
a "virus" paradigm for implementation).

BTW: the only reason that this was possible was LISP, which required
the ability to write crap out to the end of an executable and so
the OS didn't verify the executable size vs. actual file size.  On
a properly secured system, the linker would be a priviledged
program and would write an MD5 checksum or other such garbage (or
a simple data stamp) that was compared prior to loading the binary
to make it runnable.


That's a relatively simple fix if your OS is protected mode and
you have to present credentials to cross a protection domain to
write the stamp/CRC/checksum field.


It is simply statistically unlikely that you will have virus
infections on a UNIX box, mostly because it would take someone
marginally intelligent to write them, whereas a DOS/Mac/Amiga
virus can be written by any idiot (there are even "cookbooks"
for them).


                                        Terry Lambert
                                        terry@cs.weber.edu
---
Any opinions in this posting are my own and not those of my present
or previous employers.