IBM's OmniFind Yahoo Edition, mentioned in IWR a couple of days ago, has to be worth a try although don't expect instant results. There's the small matter of indexing to take care of first. I decided to let its crawlers loose on the IWR blog and website and my own hard drive.

It took a few hours to index around 35,000 documents and build an index of almost 2.5 gigabytes. It works on specific file types, which means that some proprietary format files don't get indexed. And there's a size limit on the files it will index. I don't know what it is, but the first file it baulked at was about 1 gigabyte. [Update: maximum file size is 51.2MB.]

Web crawling is continuous after a delay following the initial session but the hard drive crawling has to be sanctioned by the search administrator.

The search splash screen comes up with all the usual IBM/Yahoo branding, copyright and search destination stuff. You can replace all the elements with your own. I elected for spartan simplicity:

And here's a glimpse of the start of the results:

Set up is dead easy, although it baulked at indexing my machine from the C:\ drive. I settled for making my username folder the root of the search.

Searching is snappy. At the moment (after only half a dozen searches), the average response time is 881 milliseconds. But then you'd expect this if the index is inside your own machine.

If you want to fine-tune the search experience you can import (or key in) synonyms. You can also add featured links which appear at the top of the search results in response to certain keywords being entered. For example you could associate a support URL with the words 'help', 'problem', 'helpdesk', 'faq' and 'support'.

And there's an API for the technically-minded which delivers results as Atom feeds or HTML snippets to embed into your web pages as you wish.

Best of all, it's free. As long as you don't exceed half a million documents and you stick to the local drives in your own system [Update: and network drives mapped/mounted on the search server.]

Warning (added 16/12): You have probably realised that this version does not take account of user access policies. It's for the administrator to ensure that restricted material is not indexed, or to restrict access to the search and its cache.