A running description of activity related to DragonFly BSD.
Matthew Dillon declared his intention to have REDO working for Hammer very soon. This will improve speed by lowering the number of fsync()s needed in a given period of time to flush data to disk.
He continues in a separate message talking at length about data flushing and how to implement it efficiently, with some comparisons to work in FreeBSD. The followups are worth reading, too.
Antonio Huete Jimenez wants to find anyone working on tmpfs for DragonFly; his post about it summarizes the work so far. He’s interested in working on it, in any case.
Why, BSDTalk 184 is our very own Matthew Dillon talking about all the recent changes in DragonFly, for a good half-hour! I’ve listened to about half of it so far… I hadn’t realized the significance of some of the changes in the last two releases. It’s also strange to hear someone mentioning the work you’ve done (pkgsrc bulk builds)…
Jan Lentfer’s posted the steps to test OpenSSL encryption – I link to them because it’s interesting to see the steps spelled out.
Jan’s also posted a patch to enable DNSSEC support throughout BIND and related tools. Test if this interests you. (and it should.)
I’ve always said you can’t be too rich, too thin, or have too much RAM. (I’m paraphrasing a quote from the Dutchess of Windsor.) However, maybe you can have too much RAM. Recent changes by Matthew Dillon have made it possible to run the kernel_map out of RAM depending on the quantity of video RAM and system RAM in use.
This isn’t a significant danger; I’m highlighting it because it’s an odd problem. It’s easy to work around for now. There’s a new utility, kmapinfo, to show mow much kernel memory is being used.
In response to a question about fine-grained vs. coarse locking, Matthew Dillon detailed the locking types used by DragonFly and the remaining work left to make the system able to function completely without the Giant Lock. (hint: VM, something Matt’s known for.)
There were some errors with the dragonflybsd.org domain, which are now fixed. This includes some issues in NNTP access to the discussion groups, which is why I don’t have a link for this.
As previously noted, some of Matthew Dillon’s work was potentially destabilizing. The “danger period” (I saw no issues reported) is over.
It’s New Year’s Eve Eve, and so here are a bunch of links I’ve built up over the past few days.
BSDTalk was recently linked here interviewing Randal Schwartz. Randal Schwartz and Leo Laporte, who create a podcast called “FLOSS 101 Weekly”, now have an interview with Scott Ullrich and Chris Buechler about pfSense. (via) It’s a nice bit of symmetry, and Scott was an early contributor to DragonFly – specifically, the installer.
Everything that _why the lucky stiff did. (via) _why is one of those things that only the Internet lets exist. And he used DragonFly!
Roguelike games, evaluated via the Berlin Interpretation, on @Play. Also, a dedicated Roguelike handheld?
Naoya Sugioka is working on bringing tmpfs to DragonFly – I am a big fan of that idea.
top now uses CTIME, not WCPU.
BSDTalk (4 years old!) has 24 minutes of talk with Randal Schwartz, talking about a whole pile of different subjects. I met Randal before – he’s a decent guy.
A recruiter found me through my administrative role for DragonFly in Google’s Summer of Code, and passed along a job description. I’ll paste it after the cut. If you’re looking for a job (or know someone who might match this job), contact me and I’ll pass contact information around.
Edit: The recruiter has a similar but non-BSD job also available…)
Man, I hope this works out. In the job climate we’ve had the past year or so, helping someone get a job is very fulfilling. Plus, the job sounds cool…
I didn’t set anything up with the Digest and tumblr… Please speak up, if you did it. (found via Google)
Avalon.dragonflybsd.org was power cycled, so pkg_radd works now, as does git.dragonflybsd.org.
I love love graphs, and Alex Hornung has created a graph showing the lock contention on a DragonFly system during a buildkernel. (ganked from EFNet #dragonflybsd on IRC)
avalon.dragonflybsd.org is temporarily down, so pkg_radd will not work unless you set $BINPKG_BASE to a new mirror.
Following a link from vitunes (which has been updated), I see vimprobable, a vi-ish web browser. Interesting both for its relentless focus on keyboard controls, and for its old-style quietly angry help (from the FAQ):
How will I know if a website is genuine without a phishing filter?
Use your brain.
Also along the same lines, vimperator.
BSDTalk 182 offers a very timely 12 minutes of talk with Josh Paetzel of iXsystems, talking about the recently not-BSD then BSD-again FreeNAS. (see previous item)
Speaking of iXsystems: they have a new facility.
Alexander Polakov has suggested that nvi (what you get when you type vi) should be replaced with traditional vi, since that would deliver UTF-8 support, and nvi appears to no longer be updated. Other than one objection on split screens, I daresay everyone who needs more features treats the system vi as a fallback and has moved to a new editor. (or that)