Wednesday, April 25, 2007

RejectedByYC.com

I launched RejectedByYC.com yesterday at 3:00 AM, to a resounding thud. Response was generally negative, mostly because people didn't like the RejectedByYC name and logo. Too negative, too close to yCombinator, and basically too much of a ripoff.

So Mike and I brainstormed some ideas. Eventually we came up with Bootstrapacitor, from "Bootstrap" and "Capacitor", the idea being that we store up energy until the startup explodes in a bolt of lightning. There're also neat potential puns with the flux capacitor, which after all is shaped like a Y(Combinator). And we're not going back to the future, we're inventing the future.

Yesterday I added the stats page, and this morning I moved the announce link to the layout (for those who've already registered startups) and added Markdown support. Mike came up with a logo last night, so I may try to add that in today and adjust the layout. And we've already registered bootstrapacitor.com, so the site needs to be switched over to that domain.

We are facing an issue over time allocation, though - in many ways, Bootstrapacitor is a distraction from our main business, which is Flash game creation. Mike has made a lot of progress on the game creation engine, so we'll want to launch Diffle.com and start integrating games in with it. I'm hoping that Bootstrapacitor is something that'll speed things up and give us publicity, but if it flops again, we may be better off just concentrating on the games and pushing that out.

Thursday, April 12, 2007

yCombinator

The yCombinator responses came on Tuesday, and...we didn't get in. I have a bunch of suspicions why, but ultimately it boils down to them only being able to interview 30 groups and us not being one of them.

I was feeling pretty bummed for about an hour, then I read some of the comments on news.YC by other rejects. And I had another idea for a startup. The biggest part of yCombinator's value proposition, for us, is the ability to work in a high-energy hacker environment, where we're continually getting feedback from our peers. And that part is not limited to yCombinator: we could easily organize groups of startups together and have some friendly competition going. We just need to group startups together.

So, I'm taking a little diversion from Diffle (hopefully not more than a week or so) to work on RejectedByYC.com. It'll have forums (which I've already setup), "demo groups" of about 10-12 startups, feature announcements, quick feedback, and karma. The hope is that this'll provide motivation to many teams who were not able to get into yCombinator or other funding, letting us maintain a high level of energy even though we've been rejected. Plus, by grouping them together, we may give them a brand advantage that lone startups don't have.

PIL

Lest I fall into the forget-about-updating trap again, I'm going to post an update about the past 5 days or so. I'll enter these as 3 separate entries as they're really 3 separate topics, though in reality I'm typing them all out between compiles at work.

On Saturday, I setup PIL, the Python Imaging Library. We're using it for thumbnails. Unfortunately, PIL is a real pill to install. I must've spent close to a day on it, Googling for everything I could find and looking through several config files.

In the end, my problems were due to 3 problems:
  1. PIL requires libjpeg to operate on JPEG files. And the Debian package is not sufficient; it must be downloaded and built from source
  2. libjpeg must be installed with the "make install-lib" command. The standard make install doesn't cut it.
  3. PIL's setup.py will tell you that it successfully installed JPEG support even when it didn't.
Unfortunately, problem #3 masked problem #2, which led to a long wild-goose chase. I only found the problem when I tried a "python setup.py build_ext -i", which fails properly if libjpeg is not correctly installed. It at least told me that the header files were missing, which led me to read libjpeg's README file more carefully and fix the problem.

In my Googling, I found that many other people have had the same problem, and the PIL developers have been very unhelpful in resolving it. So once I make this public and Google indexes it, perhaps it'll prove useful to folks.

Thursday, April 5, 2007

Month's catch up

So, I haven't updated in a month. I suppose that's good, since it means I've been working instead of writing about working. Though I imagine it's rather frustrating for the readers.

I won't even try to go over the features I've implemented. Judging from the last entry, it includes, comments, comment threading, tabs/subpages for user and game pages, AJAX ratings, comment replies, private messaging, new sort orders, plenty of bugfixes, and probably some stuff I've forgotten. I'd have to go over the Subversion logs to keep track of it all.

There are two developments that I want to include, since they're not in the Subversion logs.

We got our dedicated on March 13. It's a bottom-of-the-line Celeron - no sense spending more money than we have to if it never takes off. IIRC, Mike made the order from GoDaddy on a Tuesday, we had it by Thursday night, and that's when I started setting it up. The default firewall installation blocked DNS, so we thought that the DNS hadn't propagated when it was really my own stupidity.

Even before the domain worked, I started setting it up. Made a few subdomains for dev tasks and setup Apache virtual hosts for them. We're *only* using Apache for development & administrative tasks; the main production server is Lighttpd with an SCGI connection to Python. But I didn't want to figure out PHP/Perl/WebDAV for Lighttpd and I figured it'd be good to have our tools in a separate process for security/performance reasons anyway.

Over the next couple days and the weekend, I setup user accounts, upgraded Python to 2.5 & installed libraries, installed Lighttpd and configured it, configured flup and wrote a quick launch script, learned how to write /etc/init.d scripts to start everything, and basically got things working. By the conference call on Sunday, we were mostly good. I think I needed to do a couple tweaks to SCGI and flup to get the app running, and it wasn't until the next Wednesday that I installed monit.

Speaking of which, there were a few useful utility programs we rely on heavily:
  1. Pligg, for organizing all our links. HOW-TOs, tutorials, documentation, 3rd-party download sites, potential game developers we want to approach, and competitors all go here.
  2. MediaWiki, for documentation. We aren't really using this yet.
  3. AWStats, for web server statistics. I was surprised how easy this was to setup.
  4. Munin, for monitoring server health. Ditto
  5. Monit, for automatically rebooting services when they go down.
Those were all part of the setup process on Friday/Saturday/Sunday.

The other interesting development was that we applied to yCombinator. I had previously floated the idea, but we weren't terribly interested then. However, many of our objections were answered by the ensuing blogosphere buzz (SFP applications are apparently a big deal in the startup world), so I asked my cofounders to give it a second look. Thus followed about 4 days of scrambling to get all the information we needed to fill out the application.

Then PG e-mailed us back and said we had too many weakly-committed founders, and probably too many founders in general.

I had misinterpreted the deadline, so we had an extra day to revise our app. We had an emergency conference call that night, which was probably one of the hardest and most awkward conversations I've ever had. The three founders who were not willing to commit full-time to Diffle were cool about it, though, and we ended up deciding it was okay for me and Mike to apply as a duo, and adjusted equity shares appropriately.

We're still waiting to hear back from yCombinator. Obviously, I hope we get in, since they add a lot of value. But we're going to go ahead with this anyway, even if we don't. I want to see this product exist, and we can continue to do that even if yCombinator doesn't want to invest.

Right now, Mike is working on the layout with Matt's help. I just finished adding comment replies and logout, so I've got a few cleanup items left. Then I want to move towards writing actual Flash games and starting on the game creation engine.