2008-04-04: kadmin-remctl 2.0
Not dead. Just very busy organizing my life, getting back to effectively
managing my time, cutting down on distractions, and letting myself work on
one thing at a time. All of that has meant that I've been hardly reading,
but getting other things done. This too shall pass, but hopefully onto a
small number of new things rather than back to trying to do everything at
once.
Anyway, I released a new version of kadmin-remctl, last week actually,
incorporating additional code that we needed for our guest accounts
project. In the process, I significantly restructured the configuration
and improved the capabilities for syncing with Active Directory. All of
the three supported environments (MIT Kerberos, Active Directory, and AFS
kaserver) are now optional.
And wow is this code ever ugly. I did some restructuring of that as well,
so it's easier to follow at least, but this is crying out to have a modular
structure with a set of Kerberos environment providers that you can
selectively stack. And all the running of programs via Expect is just
gross.
Oh well, time is limited, and time for making things pretty rather than
just functional is very limited. Hopefully I'll get a chance to come back
and do this right at some point.
You can get the latest version from the
kadmin-remctl
distribution page.
2008-04-04: remctl 2.12
remctl was next on the list of packages that needed a build system overhaul
to pick up my new Autoconf macros and my new LICENSE file scheme, and with
this release I switched C comment formatting styles and made that global
change as well. So a lot of code was touched here.
I also finally incorporated the Windows port by Matthew Loar. I doubtless
broke it in the process, but since I need to do another release shortly to
include the PHP and Python bindings, hopefully people will point out how I
broke it and how to fix it before that release.
There are a pile of other fixes as well. Another internal project exposed
the fact that remctl always sort of cheated about GSS-API identities and
forced them to be a Kerberos principal, which kind of works but which
disables all domain_realm mapping and host canonicalization in the GSS-API
library. This version now uses proper GSS-API host-based service names if
you don't tell it what principal to use. The server also marks all file
descriptors close-on-exec properly (and even cleans up after the Kerberos
libraries), and various other minor things got fixed.
You can get the latest version from the
remctl distribution
page.
2008-04-10: kstart 3.11
It had been exactly one year since my last release of kstart. I hadn't
realized it had been that long, although I knew I had various
user-submitted patches that I hadn't incorporated for quite some time.
kstart was one of the first (maybe the first) package that I converted to
Automake and started pushing towards my new package style, so the style has
evolved a great deal since then. I brought it fully up to date, including
the coding style, and added the beginnings of a test suite. I also redid
how the AFS system call layer works, pulling code from pam-afs-session and
putting a portability layer around it that also should work on AIX and
IRIX.
This version should fix the RHEL 5 problems we'd run into, and hopefully
should make kstart much easier to maintain going forward.
You can get the latest version from the
kstart distribution
page.
2008-04-11: Hugo book haul
It's that time of year again. I read one of the Hugo nominees already
— Charles Stross's Halting State — so four more to go.
And, of course, I had to pick up a bunch of other books at the same time.
M.M. Buckner -- War Surf (sff)
Michael Chabon -- The Yiddish Policemen's Union (sff)
Helen Cresswell -- Bagthorpes Besieged (children's)
M. John Harrison -- Nova Swing (sff)
Carolyne Larrington (trans.) -- The Poetic Edda (classic)
Barry N. Malzberg -- Breakfast in the Ruins (non-fiction)
Robert Sawyer -- Rollback (sff)
John Scalzi -- The Ghost Brigades (sff)
John Scalzi -- The Last Colony (sff)
Snorri Sturluson -- The Prose Edda (classic)
Nassim Nicholas Taleb -- Fooled by Randomness (non-fiction)
John C. Wright -- Fugitives of Chaos (sff)
I haven't been reading that much lately for one reason and another, but I'm
hoping to change that in the near future.
2008-04-16: Pod::Thread 0.10
This is the first release since 2002. I ran into another bug while
generating new INN documentation, figured out how to fix it, decided to
finally fix the bug with table of contents generation at the same time, and
released a new version. I also moved the repository to Subversion (since
I'm not quite ready to move everything to Git) and built a Debian package,
which is now in my local repository.
So far this vacation I've not been doing much on-line except working with
pictures, but I'm starting to feel the itch to do some organization and get
some things done. I've been doing very well at ignoring work, though, and
plan on continuing that.
You can get the latest version of Pod::Thread from my
web tools distribution
page.
2008-04-23: New journal
Previously, my journal was managed by Movable Type. Originally, Six Apart
seemed like a cool company, but I should have known better than to trust
anything that wasn't open source. They started screwing with the license,
it was never packaged properly, and then of course it became far too
painful to upgrade.
I've been wanting to fix this for some time, and have occasionally poked
at other blog software. Serendipity looked like a winner for a while, but
then I spent two days fighting with it and while I got it to mostly look
like what I want, that was demoralizing. I hate spending large quantities
of time fighting with someone else's software to get it to do what I want,
particularly when I then have to write plugins and additions to get what
feels to me like basic functionality. It would be one thing if I had time
and inclination to get involved as a developer, but usually what Web 2.0
software developers want to pursue and what I care about are almost
entirely at odds.
During my recent vacation, I finally realized that I just wanted something
that used all the infrastructure I already have. I have a macro language
that I like for generating web pages, I have a page design that I like;
the journal one isn't bad, but it doesn't match. I don't care about
nearly all of the features of blog software. I could use something that's
much better-designed (and free software) like ikiwiki, but I don't particularly like wikis either. (That's another
story that I should write up sometime.)
I really just wanted something that works like my book reviews work and
lets me maintain the whole thing in spin.
So I just gave up and wrote it. It probably took me more time than it
would take to beat some piece of blog software into submission, but now
it's done, I know how it works, it knows what I want to do, and I don't
have to fight with it again with every upgrade.
So this is my first post with my new journal software, with an RSS feed
generated by my own code. There will probably be bugs for a while to iron
out, but it will also make it far easier for me to maintain multiple RSS
feeds with different topics, post reviews to my journal even more
automatically, and add automatic posting for other things.
The main loss is comments. With this system, my entire web site can be
static, which I love. But that means no comment scripts. Since almost no
one was commenting on my journal anyway (as opposed to mailing me or
discussing posts elsewhere), I think I'm okay with that. Eventually I may
mirror my journal somewhere else where there are comments.
All of the old posts have been converted over, including old comments.
Old reviews with no comments have become redirects to the permanent review
pages.
Let me know if you see anything weird.
2008-04-24: kstart 3.12
Well, I figured that all the build system changes in the previous release
would have broken something. There's nothing like trying to build the
software on Solaris 8 to uncover all the edge-case build problems. There
were lots of problems with flag settings, ordering of libraries to avoid
AFS com_err, and krb5-config processing. I also implemented a suggestion
(for a different project) to verify the results of krb5-config and fixed
some command-line parsing bugs when providing a command and not using
-- first.
I can't release a version without making some functionality improvement.
For this version, I implemented Mike Horansky's excellent idea to have
krenew copy the ticket cache file when run with a command. That keeps the
ticket cache for the command isolated from the user's ticket cache, so
it's not deleted when the user logs out.
You can get the latest version from the
kstart distribution
page.
2008-04-24: wallet 0.9
I had to do one more release of kasetkey (the program included in the
wallet distribution for talking to AFS kaservers) so that we can move our
kadmin remctl interface prior to turning off the AFS kaservers. Since I
was doing a release anyway, I also implemented support for wallet store to
take data from a file or standard input, fixed a few test suite and build
issues, and added support for builddir != srcdir builds.
You can get the latest version from the
wallet distribution
page.
2008-04-25: kadmin-remctl 2.1
This is the release that goes with the wallet 0.9 release. It now uses
kasetkey for all AFS kaserver synchronization, including enable, disable,
and examine. Also included in this release is a bug fix for using the
allowed principal regex for examine and build system fixes and updates
first released with other projects.
You can get the latest version from the
kadmin-remctl
distribution page.
All posts