Riding home on the train tonight, I can see the laptop screen of the guy in front of me. As he boots Windows 2000, I glance up at his desktop picture, a huge painting of Jesus on a crucifix with the Virgin Mary kneeling and weeping in front of him at Golgotha, bracketed by two other crucified men.

Now the guy's playing Diablo II.

Feels strangely like Monday again, except that I'm extraordinarily busy today. Even my train-time isn't free from work's reach today, so I don't have much time to write.

So Dreamhost, my hosting provider, dropped a velvet brick on my head yesterday. They sent out the cheeriest email ever, talking about how they're changing all their plans to have more storage, more transfer, more domains, and such, and as an existing customer, I'm grandfathered in, getting all the new capacities for my original monthly price. Great! Whoopee! They have a link so you can see what all the new levels are. I follow it.

The one thing they didn't mention was that they now meter mySQL access using a unit they invented called a 'conuery' (I pronounce it 'con-weary' and with good reason). Where before they had unlimited mySQL access, now each database connection costs 25 'conueries' and each query costs a single 'conuery,' the logic being that establishing a database connection requires 25 times as much CPU effort as executing a query (hence conuery, 'connection-query').

On my $40/month plan I get 15 million conueries per month, which sounds like a lot but isn't really. It's enough for about 500,000 page views per month the way Fury is set up currently, and closer to 3-4 million views per month once I switch to persistent connections instead of one-off connections. That's fine, Fury typically gets about 70,000 hits a month, so that's no biggie.

The problem is Metacookie, which I host in the same account.

Though right now Metacookie's in alpha testing until I can wall off some time to bring it to beta and final release. The gist is though that for Metacookie to work, each metacookie-enabled site has a little badge graphic that is served from Metacookie's. That badge is a 'beacon,' so that when the reader requests it, my database marks them as being up-to-date on that weblog. Naturally, this requires a database query to update the user. The graphic served is only 241 bytes, but the query is now a lot more 'expensive.'

How expensive? Well, assuming that I completely solved the connection problem and only had to pay one 'point' per user-blog-view instead of the current 26 points (25 for the connection and 1 for the query), then Metacookie would only be able to handle 15 million 'beacon' hits per month (less the conueries that Fury, qwer, aoliza, randompixel and underblog will be using). Considering that there's a beacon hit every time anyone looks at the home page of any metacookie-enabled site, 15 million hits is an incredibly low number. Metafilter alone gets about 1 million hits a month, and that's just one site that might be Metacookie-enabled. After release, Metacookie will be serving to hundreds if not thousands of sites. The first impact this will have on the design is that people will have to be signed in to use Metacookie. That will reduce database activity by around 90%.

But what happens if I go over the limit? My $40/month gives me 15 million conueries, along with all the rest, 900 megs storage, 20 gigs transfer, etc. If I go over my 15 mil, it automatically charges me $5 for every million I go over. In essence, if I exceed my quota by 50%, I'm paying double my regular monthly charge. This for a database service that goes down for several hours at a time at least once a month, though this is likely the reason they're instituting limits like this.

Not to get overly technical, but the way persistent connections work is, when one of my scripts requests a database connection, the server checks if a connection with the same username, password, and host already exists, and if it does, it uses that connection instead of (ch-ching) starting a new connection (for 25 points). That's cool and great. In theory I can start one p-connection and use it for all my database stuff. This is what Dreamhost is pushing everyone to do.

Warning, database geekspeak ahead:

Here's the rub: Each mySQL server can only handle N many simultaneous connections, persistent or otherwise. The default is 16, though I don't know what Dreamhost's servers have been configured to support. Basically this means only 16 (err, n) sites or less can have a persistent query open at any given time. Try to open another and the oldest one closes to make room. Considering that each database server supports hundreds of sites, there are a lot more sites hitting the database than connection slots. The upshot is that if I'm the only one, or one of only a handful of sites using persistent connections, I'm in good shape, since the others using one-shot connections don't hold on to their slot. They take it and let it go, freeing it up for the next request. The problem is that if everyone does as Dreamhost recommends (indeed, is pushing on their wallets to do), suddenly hundreds of sites will be requesting one of 16 (err, N) connection slots, and each persistent connection will last for only a few seconds (if that), not minutes, before being automatically closed to make room for another persistent query, thus completely nullifying the theoretical benefit of using a p-connection.

Is this kind of connection-flooding likely to happen? Insufficient data. So far, though I'm waiting to hear back from them, it doesn't seem that Dreamhost provides any tools for measuring your current conuery usage, or usage-to-date this month. I suppose you just get a bill and that's how you know.

Anyhow, time to get to work. Let the barrage of 'you should host with provider X or Y' begin! :-)

Just in case the world is in want of more evidence of just how wacky Canadians can be, I present the World Rock-Paper-Scissors Society which, unlike some well-executed sites, is completely real.

It's funny to look at history and see how often, and severely, ideology got in the way of scientific discovery.

Theories of gravity, evolution, planetary motion, the spherical earth, even flight, all had to overcome ridicule and disdain, blockades not always limited to theistic beliefs.

It's so nice that now we look at science empirically. Sure, religion and morality often dictates what we should or shouldn't do (cloning, for example), but it doesn't stop new ideas from being pursued; ideas that might fly in the face of conventional scientific beliefs. Right?

Wrong.

It seems that after the Pons and Fleishman Cold Fusion debacle of 1989, researchers exploring tabletop fusion are cast into a corner smaller and darker than the one where SETI researchers cower.

Researchers exploring a new method of tabletop fusion, putting their paper through standard peer-review channels, have nevertheless been badgered at every stage, ridiculed, dismissed, and blocked wherever possible from presenting their findings.

Their paper, finally published in the peer-reviewed Science journal, almost never saw the light of day, blocked by those who believe that the concept is folly (or is already being utilized in government labs). It's a very interesting look into the 'once burned, twice shy' atmosphere of the scientific community.

Hello Everybody!

Did everyone have a good weekend? I saw Ice Age Saturday morning with Karen and Crystal, which was lots of fun and good fodder for our der rigour pickapart session over lunch afterwards (example).

We went to Borders ('your multimedia outlet if you don't have the patience for Amazon'(tm)) and I picked up a book (Derek's Design for Community), a CD (Baroque Adagios (2-disc set)), and a DVD (Say Anything).

That evening around 7pm I dropped by The Underground (the video arcade on the Berkeley campus, a whopping three blocks from my apartment) to see if they were open or if, as I suspected and was proven correct, they're closed for the week for Spring break. Walking up to Sproul Plaza, I called Gypsy's on cellphone speeddial. "Gypsy's" "Yeah, tortellini marinara?" "Okay, ten minutes." Click. It's a tradition. He recognizes my voice and knows I won't be ordering anything else. It's just nice to give them a few minutes lead time so I can get in and get out.

Anyhow, they're two blocks away and I've got 10 minutes. It's a nice night on campus and I figure I'll wander around in the night air to burn off a few minutes. On Sproul Plaza I step up to the brim of Ludwig's Fountain and walk around the circle of the fountain slowly, twice. A small part inside me feels self-conscious that someone might be watching, wondering why I would be walking around in circles, and a slightly larger part squashes that part, surprised I would actually be that self-conscious.

I step down and wander towards Sather Gate and I hear in the distance someone asking for directions to Hertz Hall. Ever helpful, I wander toward the lost traveller, and in front of Dwinnele I come across an elderly woman, around 75 years old, talking to a backpack-(the 'life on my back' kind)-laden youth, doing a lot of pointing, giving vague (and incorrect) directions, as the lady's three friends look on.

Backpack-youth looks to me for a little help and I start giving directions, but it's a little twisty and they'd get lost again. "I'm headed that way myself" I lie. "I can show you where it is."

The lady lets me know that one of her party has a cane and walks very slowly, so I might not want to walk with them. I tell her not to be silly, and we're on our way. As we take our ten-minute journey, I give them a handful of campus trivia. After eleven years I've gathered quite a bit of floatsam on most of the buildings of campus. She tells me they're here to see 'the 12-year-old Bach,' a piano prodigy. Like many of the concerts at Hertz Hall, this one is free.

My quartet drove in from Walnut Creek, from a retirement community of ten thousand, with seven thousand apartments, but no Bach prodigies.

They asked me if I'd like to join them, and I was really tempted. For all that I know about the events on campus, I've actually gone to a saddening few of them. If not for the food that even now was growing cold on the counter at Gypsy's, I would have. "Give the food away to someone else and I'll pay you later this evening" I didn't say. Instead I walked them past the Campenile, helped one of them down the steps to the bridge over Strawberry Creek, through Faculty Glade, and around to the front of Hertz Hall, giving them directions on how to get to Bancroft and down to the Zellerbach parking lot after the recital.

I'm not sure why I blew this out into a full story, but it was really nice. It was nice to walk slowly for a change. It was nice to look around and see the campus through someone else's eyes. Nice to be reminded of all the culture going on just outside my door (okay, honestly the culture going on just outside my door is something I could live without, but go a little farther and there's beauty and art to be had). I won't be in Berkeley forever. I should make the most of the remainder of the experience.

Anyhow, I had a nice dinner at home, listened to some adagios to make up for missing young Bach, and felt a little better about Berkeley.

Okay, train's pulling in to Santa Clara so I'll wrap up. I still have some notes from the Google talk at BayCHI a couple weeks ago, which I'm hoping to type up on the ride home. I've got a backlog of stuff to write about and never enough time to do the stories justice, but I'll do my best to catch up this week.

Happy Monday!

I wonder what would happen if, as a design exercise, a designer were asked to take an existing site and create the 'next generation' of that site. Flat photoshop mocks a flow chart and a sitemap.

Now take those docs to another designer, tell them this design has been around for over a year and is stale, and ask them to make the 'next generation' of the site.

Repeat through several iterations. I wonder if you'd just get different design styles, or if each designer would actually innovate, using the previous designers work as a base?

Sadly, information architects and site designers all too rarely use real metrics of demonstrated usability from existing sites when trying to revamp them. I wonder if in those cases it would be better to skip the trouble of actually building the 2nd and 3rd generation sites, and go right to the 4th, 5th, or 10th?

Of course, the real answer is to use a designer who finds out what the actual problems and successes of the previous design were, but this might make an interesting exercise, ala Photoshop Tennis.

This morning's Conductor Gary:

Good morning! Just a reminder to our passengers, this is train 523 to San Jose, not the train to Bakersfield.

That's all I'm going to say about that today. I've, umm. Well I've been told that I should ease off Bakersfield a bit. That from one of our fellow passengers, Jennifer, who was born and raise down there, and as she passed the three-tooth minimum, I'm going to respect her wishes.

Have a good ride everyone and for those keeping track, it's Monday morning.

There's a great story in today's SF Gate on Dotcom etymologies.

I woke up this morning and something felt different, yet familiar. My apartment was warmer than usual, there was more light than there should be at 6:05 am, and everything somehow felt bigger, and I was (oddly) more aware of being alone in my apartment. Not in a bad way, but just in a way I hadn't felt in years.

Hey, I just realized today's the first day of Spring. Wow.

Conductor Gary this morning:

Good morning! This is the Capitol Corridor train 523 with service to San Jose. This isn't the train to Bakersfield. If you're going to Bakersfield, this isn't the train for you, and now would be a perfect time to disembark.

I can't stress this enough, so I'll say it a different way:

If you know a woman who got her hair caught in the ceiling fan, you could be on the wrong train.

If you've ever taken the riding mower to the liquor store, you could be on the wrong train.

If you know that the toothbrush was invented in Bakersfield because if it was invented anywhere else it would have been called the 'teethbrush,' you could be on the wrong train.

Any questions? You can come on down and see me outside and we'll straighten it all out.