FAROO - Could P2P Search Change the Game?
Entrepreneurs have learned that pitching anything to investors with "we can beat Google at search" is the kiss of death. This is like pitching against Microsoft in the PC world of yesteryear. None of the high profile, well-funded search engine start-ups with cool new interfaces, social search or natural language technology have made a dent in the real world. Not even Microsoft, with its army of smart researchers and piles of cash, has been able to halt Google's relentless market share gains in search. So why do I think that a bunch of engineers in Germany at a low profile company called FAROO have a chance?
This is a company without VC backing, with a funny name, that's located far from Silicon Valley and most of their press is not even in English! The answer is: for the same reason that Microsoft's dominance was finally ended -- not by big competitors with a similar solution, but by tiny little efforts that eventually changed the game, like a young guy in Finland writing some Unix code and giving it away.
The game-changing FAROO advantage is not that it is free software, it is their Peer To Peer (P2P) architecture. This can totally change the economics of search. To quote FAROO's front page, "Copy the entire Internet to one system? Strange idea. That's what search engines try to do. Therefore they require 450,000 servers and $2 billion."
That's it. You don't need to read any further. There are a some other advantages that flow from this architecture, but the basic proposition is really that simple.
FAROO's Potential
FAROO has 3 big potential advantages and 3 major hurdles. First, the 3 big advantages. These were articulated on a site called ReviewSaurus, which is one of the few blogs paying any attention to FAROO:
"1. The search engine index is based on a real users [sic] browsing habits : That means that web index will not serve those websites on which you don’t spend time thus reducing spam results.
2. The data is not stored anywhere but your own computer and you have the control of your own data. Searching is completely anonymous.
3. As per Faroo’s plans, you’ll be able to earn money from their advertising based revenue model. This is yet not implemented, however, it’s one of their declared plans."
It is the latter point that is critical. FAROO can do this because they don't have to invest in huge server farms. The search business is about giving a good deal to both publishers and advertisers. You can offer a better deal if your costs are lower.
The "publishers" in this case are you, me, and everybody else who contributes to the growth of the FAROO index by searching and using their software. It is a real "show me the money" proposition. Imagine if every search you did put money in your pocket. By searching, you make FAROO better, so it is quite reasonable to expect a cut of the revenue. And if your search doesn't cost them much, they can share more of that revenue.
If their costs are low, they can reduce the cost of search terms for advertisers, and as Craigslist has discovered, reducing price in a zero cost environment has a dramatic effect on volumes and market share.
If they then create an API that enables entrepreneurs to build value on top...
Problems Facing FAROO
The 3 big hurdles are:
- Can P2P search scale? I am not technically competent enough to answer that question, but I have not yet seen evidence to convince me that it is impossible. The success of other P2P services such as Skype make me willing to at least reserve judgment on this issue.
- The chicken and egg problem, based on the fact that the P2P search index scales with the number of people using FAROO. When I first looked at FAROO last year, this looked like a show-stopper. As the number of searchers is very low right now, people are likely to try it, get lousy results and forget about it. When I revisited FAROO in March, I noticed that they had taken steps to address this. Their solution (not by itself a big deal and used by other alternative engines) is to aggregate search results from other engines in addition to results from FAROO. So you get some results no matter what you're looking for and thus may continue to use the software. This also scales their index faster.
- FAROO, being P2P, requires a client download. That can be an incredibly big hurdle to adoption. When I first looked at FAROO, I thought their best shot was to build on top of existing P2P services such as LimeWire or Gnutella. I now see that they have a fundamentally different strategy. They are building on top of .Net. Oops there go all the Mac and Linux heads... bye. This is potentially a serious issue as lots of early adopters tend to be on the Mac. I am a Mac user, so I could not use FAROO without borrowing somebody's PC. OSX and Linux support is "coming soon." I assume this will be via Mono. Having been so much in the Java world recently, I have no idea how well Mono works in practice.
When ReviewSaurus asked why they don't use Google for secondary results, FAROO responded in the comments:
"Google doesn’t provide a search web service API anymore (http://code.google.com/apis/soapsearch/reference.html)
The terms of service of the second Google AJAX Search API do not permit to integrate their search results into a client application.
Unfortunately not always an API is really open."
So it does look like FAROO is in the Microsoft camp - a fundamental threat to Google built on Microsoft technology, with Google not willing to play with them.
What FAROO Needs
However, FAROO needs to show the world that they are not just Microsoft acquisition-bait. That fear would alienate much of the early adopter crowd who they need to help them gain initial traction. They need some patient capital that sees this as a really big play that can eventually go public. In Europe, small, young technology firms still do go public.
FAROO needs some independence because their potential is huge. Apart from the basic economics, FAROO has really addressed the privacy issue as they explain here. It does appear that they have overcome the conflict between personalization and privacy that I wrote about in a previous post on this blog. That is huge. If users feel 100% confident that their privacy is being protected, they will reveal more and thus create a more accurate database of intentions (which will be more useful to advertisers). But that confidence is one part technical and one part institutional trust. So if they get bought by Microsoft, too many people will fear that their privacy will not be as safe as advertised.
FAROO also appears to have a simple and elegant "implicit web" approach to social search as they explain on their blog. Their key point is that the user does not have to tag or take any other explicit action to take advantage of social search:
"FAROO utilizes the implicit web to direct the crawler to places the users are interested in, to select, rank and personalize results according to the attention users paid to the content visited, and to implement behavior targeting for advertising based on present and past behavior."
At a conceptual level, combining the two Internet megatrends - search and social networks - is an obvious direction, so that your search is augmented by searches done by people in your trusted network. In practice that is tough to get right. FAROO looks like they maybe on the right track to do it, though.
Has anybody tried FAROO? Do you think P2P search can scale? Does .Net on OSX via Mono work? Let us know in the comments.
Share
Comments
Subscribe to comments for this post OR Subscribe to comments for all ReadWriteWeb posts
I just installed Faroo and gave it a go. The results were very limited. It didn't make sense to me quickly so I could find no reason to continue with it.
When i search I want a wide variety of results quickly. that's it. Google still wins.
I spent some time with the Faroo team last September and was impressed with their thoughts and vision.
it looks good but you lost me when you thought being built on .net was a bad thing.
They are building on top of .Net. Oops there go all the Mac and Linux heads... bye. This is potentially a serious issue as lots of early adopters tend to be on the Mac. I am a Mac user, so I could not use FAROO without borrowing somebody's PC. OSX and Linux support is "coming soon." I assume this will be via Mono.
Mono is pretty mature and there are plenty of programs on the mac using it.
also this line "This is potentially a serious issue as lots of early adopters tend to be on the Mac." proof and facts on this please? utter nonsense.
I think it could take off if it returned good results and people got the concept. You also missed a massive hurdle that they have not addressed. Spam results, whats to stop spammers from hooking up a ton of bots and just surfing spam?
To succeed, they need to be better than Google from day one.
When Google launched, they were better than Yahoo and MSN from day one. Banking on a gradual ramp up doesnt work in this case.
I looked at Faroo while I was at Kosmix. My biggest question for them is: why? On the results side, is there any advantage I can get with P2P that Google can't duplicate? I can't see any. And, for an average user, what's the real advantage of running a client on their machine? Moore's Law has solved the problem of massive storage and computing. We don't need search engines on our computers to get results at lightning speed.
Of all the points Bernard listed, I think the biggest obstacle is that they require a download. Startups need to realize that the age of the download is dead. Skype was the end of an era. People are far too mobile and use too many computers to worry about specific downloads anymore. Look at Joost-- they've been forced to move to a pure browser model because nobody took the download plunge. The only people who can successfully get downloads distributed are people who already have wide distribution (MSFT, Adobe, etc.) Everybody else needs to keep repeating the mantra, "The Internet is the Platform".
I think their biggest advantage is sharing the income. Everybody knows that on the Internet the most serious cash cow right now is displaying search related advertisement. Basically you and I generating money for Google, and we dont get any piece of that when we are searching.
The problem however is clearly in the architecture of their solution. It cant be good while there is not enough computer runnning it (it is similar in some extent to wikia search). A possible solution would be to involve universities or other parties with a lot of computers, who would be interested in to earn some money from such a partnership.
No Mac, No Linux? Then not for me.
Do these people not get it? Google's success isn't about search - it's about a marketing company that helps entrepreneurs make money. Their advertising is the money maker - that's why sites GAME google, to make more money, and thats why they have such good search results.
Its a MARKETING engine.
Take a look at the top results now. Brand name/trademark takes first place, good SEO places top 10, some contextual advertisements along the right, and a few links from wikipedia. Isn't that all google is these days?
Thank you very much for your profound and in-depth review of FAROO. I really read it with great pleasure.
As for the API, basically there is already one. FAROO works as a local web service returning all results as XML. All the transformation to Html is done later in the browser.
It is not yet documented, because we first need to implement some protection for FAROOs p2p principle, which requires a certain balance of usage and contribution.
We will allow third party programs to use the API both commercially and non-commercially, in mash-ups, with own interfaces.
We are aware that many of the early adopters are using either OSX or Linux. On the other hand the majority in the mass market is still using Windows.
We are currently working on a lightweight Ajax client, which will work even on the iPhone.
We could not build on top of existing Gnutella networks for several reasons:
FAROO implements a distributed storage to make sure that the information is still available in the network, even if the originating peers are leaving.
In Gnutella the information (files) stays at their originating peers. If that information is rare, it disappears from the gnutella network, if the originating peers are leaving. Gnutella requires that either all peers are searched which results in a huge traffic and response time or to implement a artificial horizon of sight, which results in incomplete results.
@ Alan: would be nice to see you again at web 2.0 expo in SF :-)
@ Darren: We are aware of the fact that for every search engine spam is a major issue.
Spam fighting is done mainly by ranking, not index filtering. This addresses consistently also the levels of gray, which are between spam and high quality content.
While we use attention for ranking, we still use link analysis for trust. Additionally, by statistical sampling we can keep our attention based ranking unbiased, even if we would have 50% rogue peers.
@ Samidth: There have been the SecondLive client, the StumbleUpon toolbar and Oovoo after Skype, to name a few. With the bandwidth rise downloads are not that uncomfortable anymore.
You are saying "The Internet is the Platform". So do we, referring to the decentralized structure of the internet, including users as part of the internet, not only reducing it to a bunch of central servers ;-)
I believe that people will jump over this "download hurdle" if they get something in return, which they can't get in a centralized system.
This might be privacy protected personalization, this might be attention based ranking, and this might be revenue sharing based on the saved infrastructure costs.
good to hear wolf, have you guys looked at creating an AIR or client or a firefox addon to do it?
Wolf, would your Ajax client overcome need for .Net? Would this have the same basic functionality?
@Darren: Microsoft Silverlight fits better into our .NET based technology than Adobe AIR. The latter might be interesting because of its data persistance layer.
We are still observing how both technologies are received by the users.
We try to be browser independend/neutral. Do you think a firefox addon is not considered as a download? Might be true.
@Bernard: In the first step only partially. The first version of the Ajax client allows to search within the distributed index, but it can't turn your computer to an active part of the distributed index. So this solution is good, if it’s used by a certain percentage of the users, e.g. on mobile devices.
Later we will provide a full featured Ajax client by using persistent storage, e.g. according to the WhatWG specification: http://www.whatwg.org/specs/web-apps/current-work/#storage
As technology evolves (Silverlight, Air, Whatwg) using FAROO will become smoother for the user.
While I don't like the fact that it is currently only available for Windows I don't see how you can fault the creators for going with Windows first.. Windows has the largest market share by far so it is natural that they would choose it as the first platform to support, no business model for this type of thing would make sense if you chose to shun Windows and initially only support Linux which barely registers in terms of the number of users compared to Windows. For now I will just have to boot up my Windows Vista virtual machine using Vmware and use that to run it, I can't wait for it to come out in Linux though.
1 - written in dot net
2 - have to download
3 - people will have to use it first in order to it become useful
4 - spammers will love it
therefore, God Exist!