Thanks to an invitation from Jimmy Wales, I was able to occupy myself a little with the alpha version of Search Wikia in the last few days. Owing to the appeal not to blog about it until Search Wikia started its open beta this Monday, I can only publish this post now. After Wales announced to start up his own search engine about a year ago, he has caused for a massive worldwide press echo. Back when it was still known as “Wikiasari” many already announced the coming demise of Google. It has taken more than a year to develop a showcaseable beta version, time to have a look at the prospective Google killer.

first impressions
Search Wikia's homepage is similarly tidy as compared to Google, it seems that here they took definite notes from the market leader: one input field as well as the “Go”-button. After we input the search query, we are led to the resultspage (SERPs „Suchmaschinenoptimierung“). As usual the first ten hits are immediately shown, more are dynamically reloaded on demand – a solution that I have seen Microsoft use before. This quickly shows that even the “alpha” designation is, in truth, just an euphemism – quality as well as quantity of results are markedly behind many searchengines that are done as a hobby. Out of courtesy I will foresee a comparison with Microsoft, Yahoo or even Google. About 10.000 hits for the query “Suchmaschienenoptimierung” (Google has 600.000) as well as such a site on the first rank remind more of Fireball's successful keywordstuffing in the last millennium than the up and coming Google-challenger 2008. What about the social component that so much hope was based on? It basically consists of a simple social network: provide personal information, link with friends and upload pictures. Once you are logged in you can discuss results, write a short text above the first hit, called a “mini”, in the internal wikisystem and be shown with your profile-photo on the right side of the results. A means to affect the results is nonexistent. That is it, really.

the technology
Search Wikia builds on two well known projects to compile its searchindex. Wikia is trying to distribute the crawling process to many voluntary machines of their comrades-in-arms through the purchase of the distributed-crawling-technology “Grub” from Looksmart for 50.000 USD in the middle of the year. This is a solution that has failed to work in the past and, I believe, will not work in this case either. On the one hand, the bandwidth is not a limiting cost-factor anymore nowadays to justify the considerably elevated administrative burden. On the other hand, this will open the gates for potential manipulation – cloaking during the data transmission to the searchengine is much more convenient than what is used now. They also characteristically fail to pay attention and exchange the Grub-Useragent for a new and personal one. This leads to Search Wikia's access being seen as “Grub/2.0 (Grub.org crawler; http://www.grub.org/; bot@grub.org)”. The information that the Grub-clients gather is then put into a format that Nutch, an open search engine based on Lucene, can read and search. Nutch's basic features can even now still be seen in every nook and corner of the searchengine. It seems that not even the default-ranking-factors of Nutch were worked on – for both Nutch as well as Search Wikia the ranking advantage of having the keyword in the URL is too strong. The cache version as well as the ranking “explanation” are also carried over firsthand.

How do you want to judge the present results? You could benevolently attest Search Wikia a large potential for improvement. If we recall though that it took the developers a whole year to come up with the present version, I do not want to extrapolate how long we can count on waiting for a version that can persist against genuine competition. Maybe the founder would have done well to take his time and come up with a real alternative out of the eye of the demanding media instead of heralding Google's downfall over a year ago – I do not expect Search Wikia to be a “Google Killer” any time soon.