I'm one of the engineers here at Lightcrest - I have a question for the Y Combinator community. We feel that the current search options for developers, publishers, and store-owners are limited and lacking features. Google eCommerce search is expensive ($25k+/yr) and Google web search is limited in scope, dumps ads, and doesn't provide nice features like typeahead. Google also doesn't provide aggressive indexing - even their new indexing service can take up to 24 hours to index new content. Other services like websolr and indextank have some nice options, but they don't let you go from 0-60 very quickly - you need to code up solutions using client libraries before you can get started, and there is no advanced crawling capabilities. E-commerce search solutions should let you rank and promote items on the fly. And you should have access to this kind of functionality for less than $25k/yr.<p>Would you use an initially free, eventually pay-as-you-go service that is substantially cheaper than Google's to address the aforementioned needs?
Hey there, Nick from Websolr here.<p>Of course I'm a bit biased here, but where search is concerned, Solr is a really great option for you. Certainly I don't need to rehash all its features and benefits here, when you could just as easily figure that out with a bit of light research :)<p>It's a very fair point that getting started with Solr has a bit of a learning curve to it. With Websolr we (fizx and I) think we at least have taken the hosting and configuration side of that off the table. But there are still client libraries to learn, and schemas to configure. I think learning some of those basics is a great investment that will pay off, but the up front cost is certainly there.<p>A word about crawling: it's definitely doable with Solr. The advanced approach is to use Nutch[1] to fetch and feed content to Solr. The simpler way is to use wget. In fact, I've got a basic shell script[2] to show off the latter :)<p>I'm always game to talk about how Solr can see more general usage, particularly when it comes to lowering the cost of getting started. So feel free to keep the questions/feedback/brainstorming coming.<p>1. <a href="http://nutch.apache.org/" rel="nofollow">http://nutch.apache.org/</a><p>2. <a href="https://gist.github.com/774946" rel="nofollow">https://gist.github.com/774946</a>
I used to work at a comparison search engine that basically didn't have the resources to devote to searching. We kept pushing for more resources allocated to this, but we just didn't have the bandwidth and/or capable engineers that were really experts at information retrieval.<p>Another engineer and I attempted to get search working right, but the difficulty when it comes to search is overwhelming and just not an easy engineering feat. Advanced search integration for integral user experience features, such as understanding user intention, query disambiguation, search term boosting, information metadata, typeahead, etc were very difficult to get right, hard to really test, and required lots of knowledge about information retrieval -- which a domain that not many engineers are familiar with.<p>I think a service that provides an easy to use API and allows the ability for companies to easily and effortlessly address some of the concerns above is always in need.
not sure. part of the benefit of the use of client libraries is that I can make sure private data stays private - 'crawlable' content is going to make me have to format it in such a way as to make sense to the crawler, and deal with making sure it's not accessible.<p>do you have a way around that? or am I missing something?