2007-02-01: Scattered
The story of this week seems to be getting distracted from one thing I was
doing to do something else entirely. Today, it was tracking down another
problem with our local Kerberos packages that caused the AFS integration to
not work on 64-bit systems.
That's fixed now, and I still managed to finish the audit script except for
the documentation, which I can write tomorrow morning. I thought I was
going to get more than that done this week, but my original ten hour
estimate for the audit script turned out to be almost exactly right.
Maybe I'm getting better at this whole time estimate thing.
Another book review tomorrow.
2007-02-03: remctl 2.5
The driving factor behind this release was the inability to start Puppet
properly via earlier versions of remctl. Since we're using Puppet to do
configuration management on our systems, that's not horribly helpful.
remctld was closing standard input, which was somehow confusing one of the
Ruby libraries that Puppet uses. Now, it opens standard input on /dev/null
instead, which should be friendlier.
In the process, though, I decided to knock some of the other things off the
to-do list and ended up adding support for unlimited length commands to the
client library. That was way hairier than I expected. The protocol is
unfortunately rather grody. I couldn't come up with a better one. Thank
heavens for test suites; without that, I could never be sure I had it
right.
I also cleaned up the protocol around maximum token sizes, enforced that
everywhere, tightened down the protocol checking in a few places, took care
of some memory leaks, and fixed a few other minor bugs.
You can get the latest version from the
remctl distribution
page.
2007-02-03: remctl 2.6
Well, sooner or later since I started being more public about remctl there
was going to be some obvious and embarassing security vulnerability.
In testing new stanford-server Debian packages this evening and deploying a
new remctl backend feature, I discovered that if the ACL file referred to
by a configuration line doesn't exist, the command is accepted.
Non-existent files and other read errors are treated as success. Worse, I
have no idea how I managed to write the code that does this. The fix is a
one-line patch, replacing a test that makes no sense with the obviously
correct one.
This bug has been around since at least 2.0, and probably since 1.11 when
include files for ACLs were first supported. Now, at least, this is tested
explicitly in the test suite.
You can get the latest version from the
remctl distribution
page. Obviously, I would encourage everyone to upgrade, as this problem
is something of a ticking time bomb. Having an ACL not be readable or not
exist is a very easy mistake to make.
2007-02-07: First 2007 haul
I've been very good about not buying too many books, but it's getting close
to time to start reading the Hugo nominees and there were new books out by
favorite authors, so it was about time for an order.
Robert Borski -- The Long and the Short of It
(non-fiction)
Neil Gaiman & Charles Vess -- Stardust (graphic
novel)
Guy Gavriel Kay -- Ysabel (sff)
C.E. Murphy -- Urban Shaman (sff)
Charles Stross -- Glasshouse (sff)
Charles Stross -- The Jennifer Morgue (sff)
I already had a copy of Stardust as a paperback, but this is the
illustrated version that actually won the World Fantasy Award and I wanted
to re-read it in its preferred form.
I expect to see one or both of the Stross's up for a Hugo this year, along
with Rainbows End by Vernor Vinge, which I already have. Beyond
that, I'm not sure what will be nominated. His Majesty's Dragon
seems likely, and I've already read and reviewed that (and wouldn't give it
a Hugo). Farthing is probably an outside shot, being pure
alternative history; it will have a much better shot at a Sideways. I'd
not object to Blood and Iron winning, but I think it's pretty
unlikely, as is Carnival.
Looking at the NESFA site, other possibilities are Eifelheim by
Michael F. Flynn (hadn't even heard of this one, but I'm behind on
Locus reading) and The Ghost Brigades by Scalzi, a sequel to
the Hugo-nominated Old Man's War. From what I've heard of the
latter, I don't think it's deserving, but I've not read it yet.
I'd like to see Sun of Suns make it, just because Karl Schroeder
deserved a Hugo for Lady of Mazes and didn't get one, but I think
it's a very outside shot.
Anyway.
The Borski is another critical analysis of the works of Gene Wolfe. This
one I should probably hold off reading until after I finish The Book
of the Long Sun and The Book of the Short Sun. Guy Gavriel Kay, of
course, is one of the authors whose works I buy immediately and read close
to immediately. Finally, C.E. Murphy is an acquaintance of a friend of
mine and I like the mystery/fantasy crossover genre, so I thought I'd give
it a shot.
2007-02-10: cvslog 1.55
Another cut at the cvsweb improvements. Now, the links to the latest
revision for new files don't try to create a diff or specify diff options,
so they should actually work. Show the previous revision for deleted files
as well, now that we've worked out the mechanics for how to do so.
Thanks again to Slava Girshman for the feedback.
You can get the latest version from the
cvslog distribution
page.
2007-02-10: fsr 1.10
Jeffrey Altman found that fsr won't work with ActiveState Perl because the
fork emulation isn't up to snuff, so fsr now dies immediately if run with
the native Windows Perl. It also falls back on fs in the user's path
instead of a hard-coded Stanford path and supports examine as a command.
You can get the latest version from the
fsr
distribution page.
2007-02-12: More time management
This weekend, I started reading David Allen's Getting Things Done
again, and ended up re-reading the whole book in a weekend. I reviewed it
when I originally read it, but I wholeheartedly recommend it again. Even
though the tone is more for marketing executives and CEOs, once I managed
to analyze what it was saying and apply it to my job, I've found it
endlessly useful. And reading it twice, separated by some months of trying
techniques, helped a lot.
Now I think I have more of a grasp on why my next action lists were getting
out of control (not enough willingness to shunt projects to the "Maybe
Later" pile and out of the active set). I was making insufficient use of
the concept of a project to group things together, and I needed a few more
location categories for tasks than I had. In particular, I need separate
categories for high-energy tasks and low-energy tasks so that I can more
easily locate tasks appropriate for having little brain. I've also been
doing a half-assed job at really collecting inputs first and then acting on
them, and always deciding what to do with something when it comes in rather
than reprocessing it multiple times.
So, today I did a huge reorganization of to-do lists, processed a bunch
more ideas that I wrote down while reading (that always happens when I read
time-management books), and started doing a better job of collecting. Too
many meetings today, so I didn't manage to finish, but I should be able to
tomorrow. We'll see if I get just a temporary spike or if I can leverage
this into another lasting improvement.
2007-02-13: cvslog 1.56
The last change didn't put newlines at the end of every cvsweb URL. Doh.
Thanks to Slava Girshman again for the report. While I was in the code, I
also limited the size of diffs to 200KB by default (the same as I did for
svnlog) and then added a command-line option to change the size cap.
You can get the latest version from the
cvslog distribution
page.
All posts