FreeBSD 10.0

[Posted January 20, 2014 by ris]

The FreeBSD Release Engineering Team has announced the release of FreeBSD 10.0. Some highlights include; clang has replaced GCC as the default compiler on some architectures, Unbound has been imported as the local caching DNS resolver, BIND has been removed, make has been replaced with bmake, ZFS improvements, major enhancements in virtualization, and more. See the release notes for more information.

FreeBSD 10.0

Posted Jan 20, 2014 21:19 UTC (Mon) by smoogen (subscriber, #97) [Link] (11 responses)

A BSD without Bind just seems wrong. I mean Berkeley Internet Name Domain... I guess Sendmail is the next thing to go (if it hasn't already).

FreeBSD 10.0

Posted Jan 20, 2014 21:58 UTC (Mon) by Lukehasnoname (guest, #65152) [Link]

AFAICT Unbound is a lightweight DNS caching server that better suits most default DNS server requirements. BIND is still available as a port/pkg.

FreeBSD 10.0

Posted Jan 20, 2014 22:40 UTC (Mon) by chojrak11 (guest, #52056) [Link] (7 responses)

Long live FreeBSD! At first tests 10.0 final seems very good release! It's stable and performs very well. Clang is really fast. My first impressions are positive.

It's a pity that BIND hasn't been slaughtered a long time ago. World's changing. BIND isn't. DNS amplification attach wouldn't be so massive had they chosen (or changed at some point) better default values for recursion and access (hasn't anything to do with FreeBSD). I hope Sendmail is next on the death list. NetBSD have done that already. And I applaud killing gcc.

FreeBSD 10.0

Posted Jan 20, 2014 23:44 UTC (Mon) by mathstuf (subscriber, #69389) [Link] (4 responses)

> I applaud killing gcc

In general or just as FreeBSD's default compiler? Any specific reasons? I'd rather reasons be against 4.7 at least and not GPLv3 related; the latter caused the first and is well-known already ;) .

FreeBSD 10.0

Posted Jan 21, 2014 9:27 UTC (Tue) by ovitters (guest, #27950) [Link] (3 responses)

Clang license fits better with *BSD than gcc and is good enough? Seems enough of a reason (prefer clang, doesn't have to be about gcc). The "killing gcc" is just emotions I guess.

FreeBSD 10.0

Posted Jan 21, 2014 15:17 UTC (Tue) by mathstuf (subscriber, #69389) [Link]

Yeah, I was looking for reasons behind the emotions :) . I certainly agree that clang is useful from a licensing homogeneity argument, but that doesn't warrant "killing gcc" IMO.

FreeBSD 10.0

Posted Jan 22, 2014 16:46 UTC (Wed) by Pawlerson (guest, #74136) [Link] (1 responses)

License is not an argument, because Linux distributions have no problems with shipping BSD packages and I don't expect freebsd developers to be so childish. Good enough is also hardly an argument. I would search for their motivations somewhere else. "killing GCC" sounds funny. BSD distributions were using GCC since years, so rather than "kill it" somebody should say: "thank you very much GCC. There wouldn't be BSD distributions without you". Or at least: "our life would be much harder".

FreeBSD 10.0

Posted Jan 22, 2014 17:42 UTC (Wed) by mathstuf (subscriber, #69389) [Link]

> I don't expect freebsd developers to be so childish

They were using GCC 4.2 by default because it was the last GPLv2 release. To install musicpd (which is now using C++11) I had to also build the gcc47 port.

FreeBSD 10.0

Posted Jan 22, 2014 0:11 UTC (Wed) by wahern (subscriber, #37304) [Link]

If Unbound can learn from BIND's mistakes, why can't BIND learn from BIND's mistakes?

Arguably it can, and it has.

FreeBSD 10.0

Posted Jan 22, 2014 9:47 UTC (Wed) by hawk (subscriber, #3195) [Link]

In my experience having actually overridden the defaults is a much more common cause for this problem than the using defaults when it comes to this.

The actual defaults are (and have been for many years) to allow recursion from localnets; localhost; and normal queries from anywhere.

The vast majority of the DNS amplification nonsense seems to be based on open recursion and while localnets may be to permissive in some scenarios it's nowhere near the wide-open recursion that seems to be the typical target for abuse (and I suppose should be comparatively straightforward to track down if the amplifier and victim are in the same network).

However, I think one factor is that there is no default configuration *file* (which is not that weird as it works with the defaults) but even a lot of maintainers of packages and whatnot seem to be confused and create a file where they stick a *lot* of unnecessary stuff that best-case is redundant and worst-case really messes with things.

FreeBSD 10.0

Posted Jan 20, 2014 23:42 UTC (Mon) by mathstuf (subscriber, #69389) [Link] (1 responses)

> I guess Sendmail is the next thing to go

It's possible. There is OpenSMTPD now[1].

[1]https://lwn.net/Articles/334866/

FreeBSD 10.0

Posted Jan 21, 2014 0:30 UTC (Tue) by fperrin (subscriber, #61941) [Link]

NetBSD also replaced sendmail with postfix in 4.0 (2007).

FreeBSD 10.0

Posted Jan 21, 2014 12:29 UTC (Tue) by djzort (guest, #57189) [Link]

props to NetApp for donating bhyve to get a freebsd "kvm" like hypervisor going.

FreeBSD 10.0

Posted Jan 22, 2014 1:20 UTC (Wed) by aboutthebsds (guest, #95107) [Link] (2 responses)

Check my blog for the most accurate articles about BSD.
http://aboutthebsds.wordpress.com/

FreeBSD 10.0

Posted Jan 22, 2014 4:49 UTC (Wed) by malor (guest, #2973) [Link] (1 responses)

For other readers: that's just a troll site. Not worth your time.

FreeBSD 10.0

Posted Jan 27, 2014 2:41 UTC (Mon) by aboutthebsds (guest, #95107) [Link]

http://aboutthebsds.wordpress.com/2013/12/25/renewed-trol...