IBM has already got an enterprise-class search product in OmniFind , so why is it partnering with Yahoo to provide a free version of the same? The answer is simple: IBM wants to attack Google’s Mini search hardware in the small business marketplace.

But don’t think of IBM OmniFind Yahoo Edition as just a toy version of the real product, or a loss-leader to grab attention and direct users to an expensive upgrade path. IBM OmniFind Yahoo is a serious tool. It might be free, but it’s one lunch that looks likely to satisfy, courtesy of the capacity to index half a million documents and an ability to combine results with Yahoo-powered web searches.

The “three-click installation” promise is no hype. It really is a doddle to install – just don’t expect it to be a quick doddle. Although installation itself took 15 minutes via the Java GUI, the entire afternoon was spent in building a 3.5GB index of the 50,000 documents found on the local drives and network drives mounted on IWR’s test search server.

Be aware that user access policies will be ignored, so the administrator must take care not to index restricted data nor restrict access to the search itself (and its cache).

IBM OmniFind Yahoo failed to index some documents because of a lack of support for the file format, including such common types as JPG, MP3 and TIF. Some 200 file types are searchable (in 30 languages), including DOC, HTML, PDF, XLS, XML and ZIP.

Look out for the maximum size limit if you deal with larger documents – anything bigger than 51.2MB and OmniFind Yahoo Edition barfs. The other major size restriction comes with the total number of documents in the index, but given that this has been set at 500,000 that really shouldn’t matter.

EXTENDABILITY

Because it is built on an Apache Lucene framework and has an open URL-based Representational State Transfer (REST) API, it is easy enough for the technically minded to deliver results as Atom feeds or embedded HTML snippets within a web page. The whole system has been developed from the ground up with custom search-based application integration in mind, making it easy to create extensions to leverage the core engine with third-party applications.

Accuracy and speed are proof of the search value pudding. Surprisingly, really surprisingly, IBM OmniFind Yahoo satisfies on both counts. Not only did it return relevant and plentiful results, it was also lightning quick in bringing them up. Compared to off-the-shelf SQL query-based software, this is light years ahead.

While the default interface is the familiar one, it is easy enough to rebrand by removing the Yahoo links, logos and colour schemes and replacing them with your own. It is just as easy to add your own synonyms to the search engine, either importing them or adding them manually, as well as creating your own sponsored links in Google style so that certain keywords always return an associated URL at the top of the search results.

The lack of multiple index support is disappointing, as you can have only a single collection of documents rather than using a single engine to index multiple sites as separate entities. But then IBM OmniFind Yahoo is clearly intended to be an entry-level product, rather than a rival to the mid-tier players. Nonetheless, Google must be quaking in its boots at the prospect of a totally free solution that outperforms its own rather expensive Mini entry-level product.

With all this ease of use and flexible familiarity it is easy to forget that IBM OmniFind Yahoo is a serious enterprise-level tool. You will need to deploy this optimised enterprise search server on server-class hardware, and use all available resources to speed up the crawl rate while delivering quick responses on multiple simultaneous queries from a myriad users.

The ease of installation, administration and use means there is no need to spend a fortune on support “wine” to accompany this free lunch. It is there if you want to imbibe (£1,369 excluding VAT per year per server), but seeing as the free online support forums are so good there appears to be little point.

What does take the edge off the appetite a little is that IBM OmniFind Yahoo is only really “free” if you happen to have the relevant server hardware with spare capacity available. Unlike the Google Mini, it is not a hardware-and-all plug-and-play solution. Factor in the cost of a server with a couple of fast processors, enough RAM and mirrored drive space, and the cost comparison becomes rather more complex.