zillablog

This has been going around the net a little bit, but for those who haven't seen it, check out the new DBA oriented webcomic "The Adventures of Ace, DBA". The strip is Oracle themed, but I suspect if you have ever had to maintain any RDBMS, you will be able to relate.

Just some quick notes on things that happened @ oscon 2008.

1) Synopsis of news stories for me looks something like Dave->Selena->Peter->Josh->Jonah->Bernier->Certification->Drizzle->Monty->Portland->Selena. (Hmm, I wonder if all of these are public)

2) Keynotes are back. A lot of people have been skipping keynote talks lately, this years crop was one of the best.

3) It's important to get something out of a talk. This year there were three talk where what I got was big red warning lights and signs to stay away. Honors go to CouchDB, LucidDB, and Damian Conway's "you can't do that in perl" talk.

4) Best meal? Honor goes to The Doug Fir again. Last year I thought that there steak might have been the best I ever had, so this year it was almost by definition not as good. Guess that means I can try the Fish next time.

5) Swag index? I think I got 4 of 6 free shirts from the conference floor, but ended up with 7 or 8 total, with some coming from after parties and similar. I'd rank this as a down year again, but given the economy, I guess it makes sense. Note I still didn't manage to nab an Ingres shirt.

6) Items I forgot to pack? Only one...pants! Don't worry, I wasn't running around the convention floor in my underoos; It's just that I packed all shorts and no jeans, which proved to be a little optimistic as the weather was a bit chilly.

7) Items accomplished from life list? I touched-down in Minnesota for the first time, though since that's essentially federal territory, I'm not sure what item that would count for. Good thing I got to see Terry Chay's talk; that's a bona-fide entry.

I always have a good time at OSCon, but this year was even better than last. Outside of PGCon, OSCon is probably the best conference for a Postgres user to go to. If you couldn't make it this year, I recommend you start planning for next.

The other night I was having dinner with a bunch of folks, and I think it was Jacob Kaplan-Moss (Django dude and Postgres user) who noted that the Postgres community's governance model was this crazy mix of distributed peer-to-peer style chaos, that on it's surface wouldn't seem functional, but somehow seems to work really well for our project (my paraphrazation). It's hard to argue. We don't have a one true leader, and we don't have a corporate master. In fact, we're so multi-faceted, we actually have two names (minimum) for the project! At times, this is a really good thing. Listening to Mark Shuttleworth's keynote last night, and his pointing out that an important part of the future of open source is having choice not just at the software level, but as users, developers, and in economic circles as well; and you're happy to note Postgres is already very much there. Of course, I promised Momjian (the elder) I'd have some angst in this post, so let me ignore the benefits of this model and focus on something a little more negative.

For those who can't wait for O'Reilly or Slideshare to get their act together, here is a OOo copy of my slides from the Pro PostgreSQL tutorial I gave this morning. Always annoying when you try multiple ways to get something out the door and you end up having to DIY. Oh well, hope everyone enjoyed the talk.

PS. Slideshare, please email sales@omniti.com and ask about how we can help you solve your current reliability problems.

UPDATE: The slides are now available from slidshare.

A couple months ago we started talking seriously about replacing CVS with a more advanced system for phpPgAdmin. We looked around a little, but ended up going with Git. We didn't have any favorites going in, though we knew moving to subversion was a waste of time, and in the end Git was probably chosen mostly because, of the distributed revision control systems out there, Git is what other projects in our sphere of influence seem to be gravitating toward.

When you read up on Git, it has a lot of awesome bells and whistles, expecially for CVS based projects. A CVS import tool, support for multiple tags/branches, setup for shared commit repositories (to match the CVS workflow), and even CVS compatible syntax and command line interface. This made Git seem like a slam dunk. Unfortunately, if you're using 3rd party project hosting services, you quickly find no one wants to set all that up for you. We use Sourceforge now, and they probably should be setting that up since they have all of the pieces already in place, but AFAIK Git isn't on the radar for them. In the end I gave up on that dream; if you're going to use git, just use Git and live with it.

There are a couple of Git providers out there, we ended up going with Github. Github, like Git itself, looks awesome up front, but has some issues going in. It's based around the RoR community, which means that using it service involves navigating unreliability and lack of documentation (since everyone should "just know" how things work), but also means you get very nice tools and a subtle influx of happiness into your development environment (adding more happiness should not be overlooked for a project where most people hack on the code in thier spare time). I should note Github offers both free and commercial accounts; right now were using just the free accounts and that's going well for us. I think some people are turned off by thier commercial side, but I think the ASP business model they have is much better than SourceForge's open source website and closed source commercial product model, so I've no bones about it.

We haven't done a release yet, though we have one on the radar, so we'll be vetting our release process against Git soon. We also haven't moved our website and project repos into Git yet either, those aren't critical for development, so we haven't felt a pressing need, but eventually that should happen too. I'd be interesting in hearing about other popular PHP projects that have moved to Git (or any DRCS system really), and if people want, I'd be happy to give more details on what we went through so far in the conversion process; I'll be at OSCon next week, so you can track me down there if you're going, or drop me an email (who knows it may even inspire additional blog posts).

It's still new and shiny (and confusing), but the more I use it, the more I like it. You can't ask for more than that.

Index pruning

Last week I ran across two different blog posts discussing removing duplicate and useless indexes from a database. Coincidentally, I have a nagging TODO item to clean up some indexes in the schema on one of the applications we have been developing. So, with inspiration in hand, here are some techniques to clean up your indexes.