Lately Twitter has been cleaning house, raising money, doing interviews and actually talking to users. In a blog post last week they did a Q&A session, directly answering questions about Twitter’s architecture.

So I have a couple of questions, too, based on a couple of discussions I’ve had with people who say they’ve seen Twitter’s architecture.

• Is it true that you only have a single master MySQL server running replication to two slaves, and the architecture doesn’t auto-switch to a hot backup when the master goes down?
• Do you really have a grand total of three physical database machines that are POWERING ALL OF TWITTER?
• Is it true that the only way you can keep Twitter alive is to have somebody sit there and watch it constantly, and then manually switch databases over and re-build when one of the slaves fail?
• Is that why most of your major outages can be traced to periods of time when former Chief Architect/server watcher Blaine Cook wasn’t there to sit and monitor the system?
• Given the record-beating outages Twitter saw in May after Cook was dismissed, is anyone there capable of keeping Twitter live?
• How long will it be until you are able to undo the damage Cook has caused to Twitter and the community?

Update: Twitter continues to be annoyingly and constructively responsive to criticism. They respond to this post here, saying “We’re working on a better architecture.” Kind of takes the air out of the balloon when you can’t get them riled up.

Our sister publication Techcrunch UK noticed that a Location services API had been added to Google Gears. The developers behind Gears have been plotting out future API additions for a while, and those plans have included having Geo-data available to mobile app developers (see the spec here). We found out today that Google is backing up their Location API with a large effort to map out cell-phone towers and wifi hotspots, so that a user’s location can be pin-pointed more precisely.

While some cell-phones have an internal GPS, the data is inaccurate indoors and not available on all devices. The other non-GPS method for accurate location data is to use the location of cell towers. Google can store the lat and long of a particular cell tower in their database, and when their software in the future sees that cell tower on a phone, they know exactly where the phone is. To boot-strap the database, both Google and Apple have been using a company called Skyhook, who drive around pin-pointing the location of cell towers. By using this method Google bypasses the need to have deals in place with network providers for positioning data. In addition to cell-phone towers, Google is also mapping out Wifi locations to form a large rogue base station almanac, which is used for both additional accuracy in location calculations, and also to point users to the nearest available access point.

Once the database has been boot-strapped with initial data and launched to developers via an API, users of the service will further refine and improve the service by having devices submit information on towers and signal strength (along with location) back to Google. This means that over time, the service improves itself and will be able to work almost anywhere in the world, regardless of local regulations, network providers or restrictions.

It is expected that the service and associated data will be made available for free to developers using Google Gears (specifically the new Windows Mobile version). For developers of mobile applications, it means that they now have a very accurate way of not only calculating a users position, but also an easy way to pinpoint other locations as a basis for a location-based service. There is also an effort to develop and define a standard API for accessing Location data and services in the browser. As with local browser storage, Google are leading the way here by implementing first and then working with other browser developers on a standard.

The Google Checkout/Ebay Paypal wars continue.

Ebay Australia currently allows merchants to accept credit cards, direct debit, money orders and checks for purchases, but from June 17 they want to allow only PayPal or cash on delivery. When the Australian Competition & Consumer Commission (ACCC) asked for public comments on the proposal a lot of of people responded. But an anonymous 38 page document that is highly critical of Ebay’s move was submitted on May 26, leading to speculation on who the author might be.

It turns out, the title of the document, hidden in the PDF metadata, gave a very good clue “Microsoft Word – 204481916_1_ACCC Submission by Google re eBay Public _2_.DOC.” An Australian named David Bromage first discovered it.

The document is still available on the ACCC’s website (and is embedded below), with the title stripped out. But the Australian newspapers are all over this now.

Google’s competing product to eBay, Google Checkout, is only available to merchants in the US and UK, so they don’t directly compete yet with PayPal in the Australian market. Apparently, that hasn’t stopped them from trying to keep their options there open.

In the document, Google says Ebay’s actions are anti-competitive, that the public benefits claimed by Ebay are “illusory” and that the proposal will result in significant public detriment. They also request that the ACCC ban Ebay from the action under the Australian Trade Practices Act.

Will eBay retaliate? Last year they temporarily pulled all Ebay advertising on Google after they announced a Google Checkout party at an Ebay event. If they get that mad over a party, I can’t imagine how they’ll respond to this 38 page treatise on the evils of PayPal.

The full document is below. And in other news, PayPal was finally able to fix that drop down menu bug that plagued users for over ten days and was ignored until the press and blogs started to pay attention.

Google Objection To Ebay AustraliaPayPal Proposal – Find Documents

Meet LaterLoop, a new bookmarking tool that was recently featured at Google’s I/O conference.

LaterLoop’s core functionality is very similar to Instapaper, which we reviewed last January. After adding either a Firefox extension or a Bookmarket to their browser, users can click “Save For Later” whenever they come across an interesting site that they don’t have time to read. These pages are saved in a list of temporary bookmarks on LaterLoop’s site, which can be accessed at a later date from a normal browser or from a mobile device (the site currently supports Blackberries, Nokia smartphones, and iPhones).

The site has also just implemented a download function that allows users to save all of their bookmarked sites into a .zip file. This will be handy for people on the go looking to catch up on their reading during a flight. Unfortunately this doesn’t work on the iPhone yet, though we can probably expect to see something once Apple’s official application store launches.

LaterLoop isn’t exactly a novel application – similar functionality can be found from toread, Firefox’s Read it Later extension, PhoneFavs, and a number of others. But it has very intuitive interface, and Google’s endorsement of the app at their I/O conference speaks volumes.

LaterLoop is the latest offering from developer Gregor Hochmuth, whose other ventures include FlickrStorm and Mento.

Katalyst Media founders Ashton Kutcher and Jason Goldberg were in Silicon Valley today. They stopped by TechCrunch HQ (aka my house) before heading off to more important meetings with Dan Rosensweig at Quadrangle Group (and the former COO of Yahoo) and YouTube cofounder Chad Hurley.

The reason for the visit? They are preparing to launch new interactive web content (with an emphasis on the interactive part) and are doing a bit of a road show to see what Silicon Valley thinks of their ideas. I had a chance to see some of the content and hear their monetization strategy. And while I can’t say much yet, this is clearly going to be really entertaining stuff. Advertisers in particular are likely to flock to the platform.

The two decided to focus on the web after a trial run with AOL that began in 2006. That partnership eventually fizzled – rumor is the content was a little too racy for AOL’s taste.

The new content isn’t just entertaining, it’s highly engaging with users and they definitely have a monetization strategy that goes beyond display and pre/post roll ads.

And Kutcher won’t be trying to lock people into interacting with content under their rules. “If people steal our stuff, it’s fantastic” he said.

Suddenly Katalyst Media is doing more than producing video, they’re now building software. That means they need to hire more developers. Their most important hire is a new CTO, Kutcher said. Prepare your resumes.

Sorry for the teaser on this. More to come.

Robert Scoble just finished a half hour interview with Twitter Founders Evan Willams and Biz Stone. The two are surprisingly candid about the scaling problems the service has had since, oh, it launched two years ago.

“The fact that people are frustrated is a sign that we built something people care about,” says Williams a few minutes in. I agree, but that’s no way to run a business. Stone also says that Twitter doubled in size in March/April this year.

Can video games save the music industry? Probably not. But video games are emerging as a powerful distribution channel for digital music downloads. Players of Grand Theft Auto IV can buy the songs they hear in the game from Amazon by making their characters dial a number on their cell phones.

Now the rock band Mötley Crüe (yes, they are still alive) is getting in on the action as well. They released a single from their latest album, Saints of Los Angeles, both in the video game Rock Band and as download on iTunes, Amazon and elsewhere. In the first week that the digital single was available for sale (the physical album won’t be released until June 24), it was downloaded 47,000 times on the Xbox alone compared to 10,000 times on iTunes and other digital download stores on the Web.

Maybe people who play Rock Band are just naturally drawn to the Crüe. Or maybe video games are just a better way to sell music. You get to hear the whole song as part of a more immersive experience. And if you just scored high in the game, you are probably more receptive to shelling out some cash for the song that helped you get there. All those feel-good endorphins have to be channeled somewhere.

Nearly a year after it was bought by Google for $100 million, FeedBurner is finally going to roll out Google’s AdSense as an advertising option for blogs and Websites that use its service to publish their feeds. FeedBurner will start with a few select publishers next week, and then expand the option to all of its customers soon afterwards.

What took them so long? That seemed to be the whole point of the acquisition.

FeedBurner intersperses ads in blog feeds between every few posts. Integrating with AdSense will allow for publishers to tap into contextual ads for their feeds, in addition to the ads that FeedBurner already sells.

Hopefully, Google also found the time to integrate its automated back-end payment system into all FeedBurner accounts. Until recently, FeedBurner was still sends out paper checks to publishers participating in its ad network. At least, that’s how TechCrunch gets paid.

Walmart has added a classified listings service to their site. Silicon Valley startup Oodle, which was founded in 2004, is powering the service.

The listings are free, which means Walmart is likely doing the deal to generate page views and advertising impressions. They also now compete with both Craigslist and eBay-owned Kijiji.

Walmart has a mixed history of success with Web businesses, but Walmart.com attracts 26 million visitors a month in the.U.S., according to comScore. Amazon attracts 47 million.

The classifieds listing’s went up quietly last week on Walmart’s site. The deal should help Oodle compeet against eBay’s Kijiji, which recently passed it in in the U.S., with 2 million unique visitors in April, versus 1.3 million for Oodle. Both trail way behind Craigslist, which has 30 million uniques, and is currently embroiled in a nasty lawsuit with eBay over Kijiji’s market entry into the U.S.

It’s nice to see blogs growing up, even if they are about babies. People.com has bought Celebrity Baby Blog, a fast-growing blog started four years ago by Danielle Friedland. She confirmed the deal earlier this week, after MediaWeek broke the story. The site has an editorial staff of 17 editors, contributors, writers, and reviewers (presumably, not all full time).

The blog is an obvious fit for People, which knows that stories about pregnant celebrities and their babies sell. (Doesn’t it seem like pregnant celebrities are on the cover of People more than anything else?). The price was not disclosed, but Friedland and staff will stay on to grow the site.

But People.com’s gain is Federated Media Publishing’s loss. With this acquisition, FM Publishing is losing yet another anchor blog from its advertising network. Last year, it lost Digg to Microsoft, and earlier this month it lost Ars Technica to Condé Nast. Now, Time Inc. (my former employer) has snapped up Celebrity Baby Blog.

Celebrity Baby is FM Publishing’s top parenting blog, and has recently started to pull in more pageviews (and thus advertising impressions) than FM stalwart BoingBoing. Since February its traffic has shot up—to 6.9 million pageviews and 720,000 unique visitors in April, according to comScore. That month, BoingBoing had more unique visitors (2 million), but fewer pageviews (3.7 million). See the chart below.

Deals like this point to the fundamental weakness of FM’s business model. When a blog in FM Publishing’s network gets big enough or gets bought, FM loses all or part of their advertising inventory. The more profitable a blog is for FM, the more likely it is to try to sell ads on its own or be taken away by a larger media company with its own ad sales force. (Disclosure: TechCrunch is also an FM partner site. They sell a portion of our ads, but we also sell our own).

That said, we hear that FM was actually very helpful in getting this deal done. It knows that its blogs can walk away at any moment (As publisher Chas Edwards told me when FM raised $50 million last month), and the only way to keep them is to deliver higher CPMs than they could otherwise get. FM also wants to be seen as the best partner for up-and-coming blogs. Generating goodwill is always a smart business practice, even if it means having to let go of a rising star.

Update: FM’s Chas Edwards got back to me. He confirms that FM helped Friedland assess the offer from Time Inc., although it did so as a favor. And although “it is not clear” what will happen to FM’s advertising relationship with Celebrity Baby Blog, he suspects that Time Inc. will take over once the current ad campaigns run out. But he says that the revenue impact of losing both Celebrity Baby Blog and the larger Ars Technica will be minimal:

We would love everybody to stay with us for life, but we realize that is not practical. In terms of a business impact, it is very minimal. No one site represents a substantial percentage of revenues.

And here are his thoughts on the importance of being a good partner, even at the end of a relationship:

I think it builds the rest of our partners’ comfort with us and the broader industry gets a better understanding that Federated Media is building almost a talent agency. We want our partners to go deep with us in a collaborative approach to building their business.

A lot of people still confuse Federated Media with an ad network. It is not just that we want to sell your ads, but we want to help you build your business and your brand. And maybe we’ll get the opportunity to participate in these exits in the future.

That’s certainly the right the attitude if he wants to keep or attract more traffic on his ad network (sorry) than will escape whenever a bigger blog graduates from FM.

If you look at YouTube’s numbers, one thing is clear: It completely dominates online video. YouTube accounts for 37 percent of all videos watched on the Internet and attracts about half of the audience, according to comScore. (And if you add in Google Video, that brings the total to 38 percent of videos watched). The No. 2 player, Fox Interactive Media (i.e., MySpace), accounts for only 4.2 percent of videos watched. And as the Forbes chart above shows, YouTube is still growing at a faster pace in terms of traffic than Google overall.

Yet when it comes to turning that market dominance into dollars, YouTube is holding back. Forbes estimates that YouTube will make $200 million in revenues this year, and $350 million next year. Although it never explains how it gets to those numbers, and they are higher than some Wall Street estimates, they are not unreasonable. (The home page alone is $175,000 a day, plus a commitment to buy $50,000 in Google ads elsewhere—that’s about $80 million a year right there. Plus each branded YouTube channel goes for $200,000. If someone from Forbes can lay out the math in comments, though, that would be helpful). Google does not break out YouTube’s revenues because, even at $200 million, it would be less than one percent of the company’s total.

A $200 million business going to $350 million is nothing to sneeze at. But if you believe eMarketer’s estimate that online video advertising will reach $1.35 billion this year, that would mean that YouTube’s share of video advertising dollars will only be 15 percent (less than half of its share of videos watched).

This gap could mean one of two things. Either YouTube is unable to make money from a large portion of its user-generated video inventory (advertisers want to stick to the home page and the safety of their own channels). Or YouTube just hasn’t turned on the money-gushing hose yet. It has built an increasingly unassailable market dominance under the shelter of Google’s wing without the need to maximize revenues. That attitude, though, is obviously changing, with YouTube now pushing AdSense for video and spreading that wealth with more content partners.

Which one is it?

b5media, a media network with over 350 blogs, has partnered with PicScout to obtain legally licensed images at no cost through their PicApp application.

PicScout originally started off as a content copyright enforcer, hunting down unlicensed images on the web with its flagship ImageTracker program. After realizing how ubiquitous unlicensed images were, the company launched PicApp, a flash-based image provider that offers legally licensed images from large databases for free. The company makes money by including ads as part of the embedded picture viewer.

b5media will use PicApp to complement a number of other image partners. Many high quality images from licensed catalogs can go for $10 and up, which can have a significant impact on a blogging network’s bottom line. The company says that using PicApp’s free content for some of these images should reduce licensing fees substantially. b5 has over 250 bloggers, which generate upwards of 10 million unique visitors monthly.

The partnership marks a big win for PicApp, but it’s hard to imagine a Flash-based image provider becoming commonplace on the web. Flash is clunky and slow compared to normal images, but at this point the plugin is a necessity for image catalogs to control their content. GumGum is another player in this space that uses similar Flash technology.

Here’s an example of a licensed image:

Image details: Cannes: Blindness – Premiere served by picapp.com

We don’t know much about Grockit. The company is creating a new way to get people to learn online, and has spent the last year working away in stealth mode. Whatever it is, it’s apparently impressing investors: Grockit just raised $8 million in Series B funding from Integral Capital and Benchmark Capital, bringing its total to $10.7 million – impressive for a product that has yet to see the light of day.

According to the company’s press release, Grockit is “a MMOLG (Massively Multi Player Online Learning Game) where people can connect to learn from each other”. The company hopes to release the product this fall.

Grockit originally launched in November 2006 as an online exam-prep class that competed with companies like Kaplan and The Princeton Review. In July 2007 Grockit announced that it had scrapped that idea in favor of their current plan, and raised a $2.7 million Series A round led by Benchmark and angel investors.

The company was founded by Farbood Nivi, who taught in the exam-prep business for years, and Michael Buffington, an experienced Rails developer.

Blaine Cook, the former Chief Architect of Twitter, takes a shot at his former company today complaining, of all things, Twitter downtime.

The only trouble is, the feature he’s complaining about hasn’t failed, it’s been taken down by Twitter along with other measures to reduce overall stress on the platform. The same platform that Blaine built and that is occasionally seen live and functioning on the Internet, that is.

I’m all for trashing Twitter when they go down (it’s turning into something of a hobby for me), but its just plain weird for Cook to be doing it. Even more so when he gets the facts wrong.

The official on the record story between Cook and Twitter is that he left on good terms, everything was amicable, needed to move to the UK for his partner’s career move, etc. The unofficial off the record story is that he was shown the door so that Twitter could get down to building something that could scale. Many more comments like this, and Twitter may fire back.

On a side note, Twitter’s doing an excellent job of actually communicating with users all of a sudden. Good for them, it was much needed.

Earlier today we wrote about a nasty survey scam that pre-IPO Intelius, the company founded by Naveen Jain after he left Infospace in late 2002, runs by customers after checkout. These customers are asked if they’d like to fill out a quick survey to get $10 cash back, but what they are really doing, via the fine print, is authorizing the transfer of their credit card information to a third party. The customer is then auto enrolled into a perpetual $20/month subscription for what appears to a completely bogus service.

We also pointed out that nearly all of Intelius’ revenue growth is driven by this survey scam, and questioned the IPO lawyers, bankers and accountants who have failed to discover it during the months-long due diligence process.

Jain’s son Ankur Jain (pictured right with Richard Branson) sent me a message via Facebook defending his father and suggesting that our original post is uninformed and misleading.

Ankur makes a number of assertions in his email. Among the more important: Ankur states that his father was cleared of any wrongdoing around the InfoSpace insider trading debacle, for example, and that the Seattle Times made false allegations against him (the three part Seattle Times story on Jain is here). Based on my review of the lawsuits, it seems that Jain was far from cleared of wrongdoing and the Seattle Times articles appear to be objective and well researched. In short, Naveen Jain did some fairly evil stuff with InfoSpace and paid out substantial amounts of money in fines and judgments.

Ankur also says that Intelius has “one of the strongest business models I have come across” and that the Adaptive Marketing survey scam/perpetual subscription is actually a legitimate service. As I described in my original post, based on my purchase of Intelius products and the subsequent voluntary tripping of the survey scam, I disagree strongly with both assertions. So do hundreds of consumers who’ve complained to the Better Business Bureau and other organizations.

The full message is below:

Read More

As Ruby on Rails devotees converge upon RailsConf 2008 (and the simultaneously held un-conference CabooseConf), performance startup FiveRuns is launching TuneUp, a “social” debugging tool for Rails applications.

The TuneUp plugin tells you specifically where a RoR app is running slowly. If you’ve coded a few ridiculously inefficient database queries, it’ll point out just which ones.

But debugging is not always so simple, so TuneUp does something sorely needed in a world dominated by Google searches for programming answers. A reporting mechanism sends your reports to TuneUp’s site for review by others. Each report, or “run”, contains your complete Rails configuration, the entire execution path of a request, and the overall execution time. When you publish these runs, other programmers and team members can diagnose your problems and offer potential fixes.

TuneUp has the potential to get rid of long posts on Ruby forums detailing programmers’ configurations, exact SQL queries, etc. Over time (if FiveRuns structures it correctly), TuneUp may form a great pool of knowledge with common programming issues and bugs – and their answers.

FiveRuns belongs in the quickly growing category of Rails applications supporting other Rails applications. New Relic, Heroku, and Engine Yard are others in the category that have recently raised big venture rounds. FiveRuns itself has raised $9 million total from Austin Ventures.

See the video below for a screencast of TuneUp.

When serial entrepreneur Naveen Jain left the company he founded, InfoSpace, in disgrace in late 2002, a lot of people thought he would never be trusted by the financial markets again (see this three part series from the Seattle Times that talks extensively about the rise and fall of Jain at Infospace and details his violations of insider trading laws). At its height Infospace was worth $31 billion. Today it’s worth less than 1% of that.

But memories are short, it seems. After leaving InfoSpace Jain started a new company, Intelius, across the street from his old offices in Bellevue, Washington. The company sells background information on people – they describe themselves as an “information commerce company.” They’ve grown rapidly and now claim that over four million people have purchased products from them. Revenue has grown from $18.1 million in 2004 to $88.5 million in 2007. In their most recent fiscal quarter, ending March 31, 2008 the company had $31.8 million in revenue, a nearly $130 million run rate. They are also very profitable, with $22.5 million in EBITDA in 2007.

It’s no surprise that the company’s revenue growth and profitability have led them to pursue an IPO. Well known investment banks Deutsche Bank and UBS are underwriting the deal, which was first filed with the SEC on January 10. The most recent version of their registration statement, filed on May 19, is here.

Given Jain’s history, you’d think he’d go out of his way to be squeaky clean at his most recent startup, particularly as the company is going public and under significant scrutiny. But that may not be the case.
Read More

Fake Shel interviews Zivity Cofounder Cyan Banister and, as usual, there’s a surprise at the end (although regular TechCrunch readers may not be so surprised). I’ve managed to stop laughing long enough to get this post up.

Note: Not fully safe for work. And in fact a lot of you may be flat out offended.

Here’s a list of upcoming conferences that may be of interest:

OReilly’s Graphing Social Patterns, June 9-10 in Washington, DC. Use “gspe08tech” for a 15% registration discount. Thanks, OReilly, for giving away two free tickets to the conference to TechCrunch readers.

Supernova, June 16-18 in San Francisco, CA. TechCrunch readers automatically receive a $200 discount here. Come join TechCrunch as we co-host the Mobile Connections forum with Kevin Werbach Monday night at the conference.

OReilly’s Velocity Conference, June 23-24 in Burlingame, CA. Use “vel08tech” for a 15% registration discount.

Thanks also to our great group of sponsors who make reading TechCrunch possible.

ScribeFire, Firefox extension for integrated blogging in your broswer

eBuddy, web-services meta instant messenger

Levelwing, Internet advertising agency

BoonEx, community software

Rackspace, hosting services

RaiseCapital, connecting entrepreneurs and investors online

MediaTemple, TechCrunch’s own hosting provider

Check out the new advertising sponsorship packages available through the TechCrunch Network:

• 50% sponsorship rotations on TechCrunch, an affordable new way to gain visibility with TechCrunch readers.
• RSS feed sponsorships. Reach over 940,000 TechCrunch RSS subscribers daily– some of our most rabid readers.
• New sponsorship units on our fast-growing CrunchBase directory.
• Bundled pricing options for participation across multiple TechCrunch Network websites, including CrunchGear, MobileCrunch, TechCrunch UK and TechCrunch FR.

Learn more here or contact Heather Harde.

Revver co-founder Oliver Luckett (on right in picture) and CTO Rob Maigret left Revver in a management shakeup in late 2006. By June 2007 they had founded their new venture, Digisynd, based in Burbank, California. In May 2008 they raised a seed round of financing from Greycroft Ventures and Villiage Ventures.

Digisynd has been deep in stealth mode, although we’ve been able to gather some basic details about them based on information from Greycroft’s site and a job listing from late 2007:

DigiSynd is an outsourced packaging, syndication, and marketing solution that enables digital studios and other content creators to get the most value out of their content online.

The company uses time-honored storytelling techniques and valuable lessons from traditional media, but applies them in a medium of online communities, virtual experiences, interactivity and user-generated content.

That’s not much to go on. But rumor is the company has sold to Disney, months before the actual launch of the product. We have this from multiple sources, although the company will not comment. Literally, in fact. When I cornered Luckett at the D6 conference yesterday and asked him about the deal he literally walked away from me. His VC, Dana Settle from Greycroft, pretended she didn’t hear my question the three times I asked her, and then also walked away. Sounds like a confirmation to me.

From what we can tell, the deal is not being done through the strike team Disney set up last year to do Internet based acquisitions, but is being led by someone else in the organization. More details as they emerge.

Credit to Brett Williams for the picture above.