On Fri, 09 Apr 2010 15:49:58 +0200, "Jan G.B." said:
And where's the point in reporting several projects that use a -say-
library which has a reported problem? (I mean, you've send quite the
same mail with a different software to bugtraq, today.)
A few years ago, a rather nasty vulnerability was found in the zlib
compression library. We then saw a whole raft of advisories for things
that included the zlib libraries, because often the package shipped with
a private copy of zlib so patching the system zlib did *not* actually
fix the problem for the zlib-using package.
And quite frankly, if it's a very low-level package, the average system
admin may not even *realize* that his very important MobyFoo package that
he remembers uses something called FooBar (or at least he remembers MobyFoo
wanting FooBar when he installed it 3 years ago), and the year after that,
FooBar started using QuuxBaz, which (a) the sysadmin didn't even know was
installed on his box, and (b) has a security hole.
You think I'm kidding? Even *after* some vigorous pruning, my Fedora laptop
has 1,782 RPMs installed - back around Red Hat 9 it was more like 600. Lotta
software bloat going on, and most sysadmins don't have the combo of time and
clue to fight it. For instance, it's a losing battle to keep Bluetooth
software off this laptop, even though it doesn't *have* Bluetooth hardware,
because more and more packages link in Bluetooth "in case you have it".
And not one of those package developers understands the concept of a linker
"weak reference". Argh.
