Daily Archives: May 31, 2008

Gillmor Gang, FriendFeed edition

Steve Gillmor had the founders of FriendFeed on the Gillmor Gang yesterday. Lots of interesting info about what they are thinking about.

Also yesterday Jeremiah Owyang also visited FriendFeed’s offices and came away with an interesting post about MicroMemes.

We’re talking about the Gillmor Gang and whether or not you think Steve Gillmor is a rude host here on FriendFeed. Listen in at minute 50 where Marc Canter gets after him for trying to tell FriendFeed what to do.

Missed news: Google underprices Amazon by 10x

This is a good point that Google’s new App Engine underprices Amazon’s Simple DB by 10 x.

Google is making moves and these are not to be missed.

UPDATE: TechCrunch is reporting that Google is shipping some new geolocation services that he thinks are a big deal.

UPDATE2: We’re discussing the pricing over on FriendFeed where someone points out that the comparison isn’t exactly fair.

UPDATE3: yesterday Google announced a health data API. I think this could be bigger than any of the other stuff we’re talking about today.

Clearing the air with Twitter

Yesterday Evan Williams (co-founder of Twitter, his Twitter account is here) wrote me an email telling me he wasn’t blaming me and trying to clear the air. I said “can I come over?” to talk more about these issues face-to-face. I’ve always found that dealing with unpleasant topics is always better face-to-face and not over email. He wrote back and said to come on over.

I wasn’t expecting to do a video interview. I hadn’t had any sleep in 30 hours. I was tired and had just finished a dress rehearsal for our new “WorkFast.tv” show that we’re filming at Revision 3 and after that I was on Gillmor Gang, who had FriendFeed’s founders on. The show isn’t up yet, but is a very interesting one.

When I arrived at Twitter Evan Williams met me at the door (I had Twittered that I was going to be there in a few minutes). We had an off-the-record conversation which wasn’t, let’s say, fun. But we both cleared the air and then the conversation started getting interesting and I asked “can I turn on my cell phone and start broadcasting this because I think a lot of people would be interested?”

They said yes and now you can watch the rest of the conversation. I’m sorry about my goofy laugh. I was nervous and tired and it gets worse when I am both of those.

Over on TechCrunch people are giving me heck for not using a professional camera. The audio is a little rough to get. I would never have gotten this interview if I dragged around my professional camera everywhere. This wasn’t an interview opportunitity. It turned into one while we were there. Professional cameras are not appropriate things to drag along everywhere you go. So more of these interviews will be ahead and the bad audio and goofy laugh you’ll just have to deal with.

I’ll spend a while this morning to put a rough transcript of what they said here. Come back and visit this post later if you want to see what they said in text without watching the video.

It’s a 27 minute long conversation.

Outline of what we discussed, not word for word: Evan Williams and Biz Stone, co-founders of Twitter. Jesse Stay, a developer from Salt Lake and a Twitter user, was along (we were supposed to have lunch).

SCOBLE: Let’s clear the air. (This part takes the first 3:15 of interview)

EV: “For the record, we are never blaming Scoble.” They then explained how Alex Payne, a developer with Twitter, wrote a post where he was trying to explain what’s going on with the service and wasn’t trying to blame anyone.

The latest thing is that one of our developers wrote a blog post last night where he was trying to explain things, be more transparent.

EV: Explained that Scoble’s use is a “power use” of the system.

BIZ: Said one key part of the post was admitting that the service was not up to par and that the team didn’t have good enough technology in place yet to deal with the loads that Twitter is seeing.

EV: At 3:15 the discussion shifts to talk of the Instant Messaging functionality being down and why the service has been so bad in the past two weeks. He notes that in past 24 hours they’ve seen 37 minutes of downtime which, while being bad and unacceptable, is an improvement.

05:00 I note that Twitter users are extremely loyal and that even after all the downtime I’m still seeing a Tweet come in every second or two. EV discusses how fortunate Twitter is to have so many users who love the service.

05:50 BIZ talks about the plan to get out of the hole. Admits that Twitter could still have significant problems for “months.”

EV: discussed how dismaying the past two weeks have been since they had months of the service being relatively stable and had even survived the load at SXSW.

7:35 EV takes on all the people who think they have the “quick answer” noting that lots of people have told him “why don’t you get rid of Ruby already.” Notes that money isn’t the problem and that the problem is an architectural problem. At 8:24 they note that they could put me on my own server, but then no one would be able to talk with me. I thought that would be funny to many of my readers.

8:50: Jesse asks why they couldn’t open source their code base and get people outside of Twitter to help out. EV: says that won’t help, and admits that they only had four engineers who hadn’t solved this kind of problem before and that they are curing that problem now.

At this point, at about 10 minutes into the conversation, a ton of people joined the conversation and started to talk to me via Qik.com’s commenting feature. One reason why I use a cell phone is to be able to get live feedback from my audience.

10:45 I ask about the new ability for them to be able to turn off pieces of the service (right now, for instance, the XMPP gateway is turned off so we can’t use IM clients with Twitter).

EV answers that they’ve had this ability since before this year’s SXSW service and that they engineered this to be able to turn off services that are causing too much load on the core Twitter system.

13:03: I ask about how Twitter’s engine works internally and I ask if Tweets are copied for each Twitter message. For instance, do my Tweets get copied 23,000 times? EV answers that the service does NOT do that. Then talks about Twitter architecture and what they’ve learned over the last two years and how that’s showing them a path to a new architecture.

19:25 Why doesn’t Twitter stop taking new users until they fix the problems. EV says that wouldn’t make much difference.

21:50 Why doesn’t Twitter stop accounts that use scripts? EV: says that’s tough to do because Twitter has APIs and that if you turn off the people who aren’t using it in cool ways you also have to turn off the people who are using the APIs properly.

22:30 Why didn’t they build Twitter right to handle all these problems from the start? EV tells us about the history of Twitter and explains that it was built “on a lark” and that no one expected it to be a big deal. I agreed, remembering how everyone told me “Twitter is lame” first time I told them about it.