laptopless …

So, for the second time in as many months, my laptop is going back for repairs.  Strange problem, another one of those things that I’m apparently not sure how to google for … either that, or it really is some kind of isolated problem.

What happens is that the backlight on my Toshiba m55 is blanking out on me … when I open the hinge past a certain point.  That’s not good, because that certain point is well below the one where I can actually read what’s on the screen without hunching over and reading over my hands … And, of course, the problem is getting worse.  It gets harder and harder to even get the backlight to turn on.  Happened to me earlier, sent it to the shop, and they swapped out the backlight.  Apparently, that wasn’t the initial problem, because whatever’s wrong worked it’s way through a second one in pretty short order.

So, it’s back in the shop.  Hopefully, this doesn’t take too long.  Otherwise, it’s back to “what’s a good laptop …”  And this one was only 10 months old; not a good record.

Virtual List Domains and Reply-To

I discovered something “interesting” about my mailing list setup and how it’s interpreted by some mail readers. When I was trying to debug it, nothing came up, so I’m blogging here to leave a record. Hopefully, the search engines manage to pick it up. This is a bit tricky, however…

The setup: I run just under a dozen mailing lists for athas.org. The server they live on is the same one that this blog, and several other domains, are hosted on. Because I’m hosting several domains, and I don’t want email bleed (because they’re not all mine), I set things up more or less as described in this howto. So, my mailserver operates off the idea of virtual domains. There’s a special virtual domain, lists.athas.org — all emails sent to this domain are piped into mailman. It’s worked out pretty cleanly, and by doing it that way, my mailing lists are well-deliniated, and I don’t have to update aliases at the mail-server level every time I add or remove a mailing list.

The problem cropped up when I got an email from one of my list subscribers who was using yahoo mail. It turns out that he was getting bounces whenever he tried to respond to a mail from the list. So, I checked out my logs, and found his bounce in my postfix logs. Turns out that he was trying to send email to mochajava.athas.org instead of lists.athas.org. Mochajava is the “real” name of the server.

I had him forward me a copy of a mail from that list. Reply-To header set correctly. So, I know what the problem is: His mail user agent was sending an email to the wrong address. I just had no idea why. Looking through the headers, the only instance of “mochajava.athas.org” I found was in the Recieved headers.

It wasn’t until a couple of weeks later that I realized what it was. lists.athas.org was registered as a CNAME alias for mochajava.athas.org in my DNS setup. So, a ns-lookup gives back something like:

Non-authoritative answer:
Name: mochajava.athas.org
Address: 65.78.231.230
Aliases: lists.athas.org

So, I think what happened was this: Yahoo mail client does a DNS lookup to ensure that the domain in the reply-to actually exists — makes sense from a user-friendly sort of approach — and, upon seeing that the given domain is an alias, replaces the name wih the host’s cannonical name. Unfortunately, my mail server’s setup depended on the use of that alias to direct emails to the list server software …

The solution was simple: Change the DNS entry. Now, lists.athas.org, instead of being a CNAME alias, is a top-level A record. Everything works smoothly now, and replys to the list from my yahoo users actually get sent to the list.

Here’s hoping this helps someone spend less time scratching their heads than I had to…

gReader

After encountering yet-another subscriptions?-what-subscriptions? bug in Rojo, I decided to take my most recent OPML export and check out google’s Reader again. I’ve heard some good things about it — after it’s recent redesign it’s supposedly actually usuable as an aggregator …

Look ma, an RSS aggregator that can go through a software update without completely tanking!

I’ve been on it for about a week. Seems to flow decently enough. I have a couple of complaints, but the fact that I can go through it using (mostly) keyboard commands is a big plus. It *mostly* fits in well with my feed reading style, which, for lack of a better term, I’ll call skim-and-queue. Skim through all my feeds, and queue up items I want to read in depty by opening them up in a new tab. Then, close the aggregator, and make my way through the tabs. Works well for me. Thanks to greader’s shortcuts, I don’t have to deal with the mouse, except to open in a new tab — while the title link works to open items in a new window, the keyboard shortcut gets caught by Firefox’s popup-blocker. Most annoying.

That’s a fairly minor thing though. I do wish I could find an aggregator that would let me specify an oldest-first sorting for my unread items though. Having to jump to the bottom in order to read sequentially is one of the really annoying things about … well, just about any aggregator.

So, it looks like, for the time being, I’ll be dropping yet another daily bit into Google’s silo … Until, I suppose, I get annoyed enough with the whole thing to write-my-own-damn-aggregator ….

Site plottings

I’m working on a rewrite of Athas.org, converting from PHP to Rails. I’m migrating off the current database, so nothing really should be lost.

Aside from the fact that it looks like I’m going to need to upgrade apache to a version that’s not packaged for my OS (2.2, because I need a mod that’s only available in that version) there are a couple of other things that I’m planning on finishing before migrating the site. After that, I’ve got a few ideas for incremental addons.

Continue reading ‘Site plottings’

npr on caffeine

NPR ran a story this morning on caffiene and coffee. ..

s/dapper/edgy/g

upgrading my laptop tonight to the (soon-to-be) beta of the next Ubuntu release. I’ll try and catalog how it goes.

At the moment, it looks like I’ve got over a gigs’ worth of pacakages to download … definately an overnight job. So, aptitude -d dist-upgrade it is…

So, a little over 1,100 packages takes about half an hour to install on my 2.something or other laptop. Not too shabby. It’s not a complete upgrade — there’s another handful of packages that had to be kept around through the first dist-upgrade. Next round looks like it’s going to take out sysvinit and replace it with upstart. Think I’m going to delay that one, as I need to head off for the moment … more later

Second round of dist-upgrades … This is interesting. The system automatically wants to install upstart … and forces the uninstallation of sysvinit, which dpkg doesn’t like. And because of the vaugeries of aptitude, I have to type “I understand that this is a very bad idea” not once, but twice, for the OS induced replacement of a critical package. Not that I wasn’t planning on doing that anyway, but, there might want to be some though given to that — it comes off a bit scary (well, replacing sysvinit should be scary, because without it, you got nothin’) and the upgrade is supposed to be “smooth” …

Despite the warnings, however, my machine rebooted quite happily. Which is good, because I hate having to dig up rescue CDs … So far, seems good. I like the new artwork. Edgy installs firefox2 beta, which means that most of the extensions I had installed won’t work … I hadn’t been too dependant on those, although I’ll definately miss the delicious plugin. Nothing earthshattering those, and those’ll come back in time.

Rojo update

I’ve been using Rojo for about a year. It’s a nice little web-based aggregator, and suits my usages better than bloglines did. Last week, they rolled out a major update to the site. And, last week, I was going to write something nice and excited about it. It seemed like they had fixed some of the speed issues, and the interface tweaks seem much more polished to me.

Clearly, I didn’t. Because after getting that first glimpse, all of my feeds vanished. Then they were back, but inconsistantly. And the only way that I could read anything was if I told it that I wanted to see all read and undread items in my feeds — not exactly the ideal way to read RSS feeds, especialy when you’re floating somewhere around the 50 or so that I track …

A week or so later, they seem to have gotten all the kinks worked out, though I’m still noticing occasional glitches. From what they say, it was a load thing — they had adapted the site to use a full-on SOA, with various bits of functionality being farmed out as web service calls. Like, for example, what feeds I read. Remarkably, I suppose, the authentication service never had a problem … Anyway, they ran a test environment, which didn’t managed to simulate their real usage loads, and the secondary servers collapsed under the full load.

I’m pretty sure there’s a lot of lessons to be pulled from this, not the least of which is that it seems they did a rather poor job of communicating what was going on, and that it would have been wise to roll back to the previous version after discovering the major (as in, broken core functionality) failures. I’m sure there are some other lessons to be gleaned from browing the reactions posted on their forums, but those strike me as two of the major ones.

Stu-ing for a couple of days …

Stu is in town for a couple of days giving talks at the NovaJUGand NoVaRUG meetings. Always good to hear him talk; he manages to convey a strong sense of the subject without drowning you in hairy details. This is exceptionally impressive when he’s talking about a subject like concurrency, which he did yesterday evening. I knew that the JVM was making some impressive strides with cocurrency, but I honestly havn’t dealt much with it since my backend days, some years back … and of course, server-side web programming has been all about the “one request, one thread” model of concurrency. A very informative catch-up though.

Nice to finally be able to quantify why I’ve been hearing for years that adding extra processor (thread)s gives you something less than idea performance growth. That being mumble’s law: 1/(F +((1-F)/N)) … looks a bit less like LISP when you’ve got a proper equation formatter, but the effects of F (non-concurrent parts of your codebase) are pretty substantial.

Tonight, he’s going to be talking about streamlined which should be interesting. I absolutely love the tagline: “Stop banging rocks together and build something.”

Dragon article

I just finished up and submitted an article to Dragon Magazine. It’s another Dark Sun article, for the January Campaign Classics issue. The article is intended for a planescape style “travel to another world” campaign, which pretty much focuses you in some awkward directions — especially for a world that’s always wanted nothing to do with the standard D&D cosmology. Think I did pretty well with it though. I’ll see if the editor gets back to me.

I owe Brax a thanks for looking over the article for me, doing a bit of editing, and shaving off about 250 words, all on short notice.

demo fallout

Well, the demo a couple of days ago went smashingly. This was the good sort of “stunning him speechless”. Now, we get a bit of rest, though apparently we’re in for a wave of new mini-demos, now that the control gate has passed and we’re on the path. Looks like we’ll need to throw a branch into the svn repository… :D