Date: Mon, 11 Dec 2017 12:14:55 -0800 From: "Chris H" <portmaster@BSDforge.com> To: "Kurt Jaeger" <lists@opsec.eu> Cc: <freebsd-ports@freebsd.org> Subject: Re: Procmail Vulnerabilities check Message-ID: <5fd8da61ee9e3a818fbb43f5e461d054@udns.ultimatedns.net> In-Reply-To: <20171211183649.GB2827@home.opsec.eu>
next in thread | previous in thread | raw e-mail | index | archive | help
On Mon, 11 Dec 2017 19:36:49 +0100 "Kurt Jaeger" <lists@opsec=2Eeu> said > Hi! >=20 > > if the majority of people install their systems via packages, that make= s for > > a fairly common FreeBSD base across all users=2E >=20 > Why would a system installed via packaged be more homogenous than > one installed as base, and updated via freebsd-update ? I don't > understand this -- can you elaborate ? OK=2E I'll try=2E I'm afraid I sort of went on a Jag, and didn't really make a = good point -- if *any* point=2E Sorry=2E But to the point, and sorry for the (additional) deviation; If I have a user base that shares a near identical install=2E I am far closer to finding/having a pattern I can work with to *exploit*, as an evil hacker= =2E So here's the thing; working from the history of Linux, and for that matter= , even MS products=2E=2E=2E someone discovers an exploit in FreeBSD, or some compon= ent common to FreeBSD=2E I can take down a *much* greater number of users, now th= at the (larger) portion of FreeBSD' user base share such a common install base= -- applications(ports)/kernel et al; are pretty much all the same for *everyon= e* because of the introduction of pkg(8)=2E Yes=2E But what's the difference if they made everything from ports(7)? IMHO, and experience, users confronted with options during build time, are *more* likely to actually *choose* options that better suite their use/need= s=2E But using packages is easier, and so if in the end everything just *works*=2E There's little incentive to use that scary "make" thing, and have to learn all those intimidating things associated with the ports system=2E Well, FLAVORS should solve all that=2E Wouldn't it? That *does* seem like a strong argument, and while I applaud all the effort= s, and those that are responsible for those efforts=2E The jury is still out=2E FLAVORS has yet to *fully* arrive=2E So it's just too early to say for sure=2E But I would agree that it *should*=2E When I look back at all the security threats that Linux had to deal with (even now), and how the ultimate argument was so often; use *BSD, it's a much more secure OS by design=2E Which was true=2E Linux was/is always installe= d in packages, or by what ever moniker they use for them=2E With that, and thei= r choice of kernel arrangement=2E They were left as easier targets than the BSD family of operating systems=2E Now looking at the increasingly narrowing of differences between the two=2E I can't help but think that the threat vector gap is *also* narrowing=2E >=20 > > In closing, and more to the point regarding Sendmail; Sendmail has a ne= arly > > impeccable security record in at the last decade=2E It provides a *secure= *, > > more powerful, and more flexible MX on the cheap=2E I see little reason t= o > > consider it an attack vector=2E Which makes *security*, and it's related > > maintenance a pretty poor argument, for it's removal=2E >=20 > The argument is: The update process for base is more complex > than for packages, and we've come a long way to have a very > nice pkg-system, in general=2E The mid-term plan is thus to package base, t= oo=2E >=20 > Packaging base means sensible packages have to be defined, and > sendmail suits a package very well=2E Indeed it *does*, and *should* be a package installed *along* with $BASE=2E That's my only argument there=2E :-) Thanks for your thoughtful reply, Kurt! --Chris >=20 > --=20 > pi@opsec=2Eeu +49 171 3101372 3 years to= go > !
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?5fd8da61ee9e3a818fbb43f5e461d054>