| Review of Inktomi Search/Enterprise|
Summary: very easy, very fast, very cool, very expensive
Since the emrergence of Google, the demand for
search engine software has dwindled rapidly. This is
unfortunate because the world needs multiple players.
Anyway, here are my notes on the Inktomi Search.
The installation of the Inktomi Search (formerly known as UltraSeek Server), I tried the Windows NT version, the engine itself is
written in Java, hence portable), could not have been
smoother. The process of creating a catalog, populating the index
(I chose directory crawl; others are database fetch,
http crawl, and API controlled add) and implementing a prototype
searching were very easy.
The default search results interface was clean and crisp.
Rich Feature Set
I played with the smartness of the search algorithm and found it pretty satisfactory. I wanted to define rules for tuning the search results (you know rules like "if the keyword appears in the teaser it is more important than if it appears in the body", or "if my site has it, it is
more relevant than an external site"), but I was unable to it. I read through mountains of documentation including APIs to no avail. I am sure there is a way, but neither the manual nor the support could tell me in a precise way.
I also explored the integration of rights checking with search results -- you know "can this user see this document?". Implementing this feature did not seem straight forward. This was the only area where
Inktomi suffered against Microsoft SQL Full-text search engine, which uses a SQL-like language to search the catalog.
One feature of the Inktomi indexer that appealed to me was the ability to break the full text document into tags. Let's say you want to catalog an electronic book, and that it should be searchable by Title, Author, Abstracts and Full text. All other search systems I know of require that I store these fields seperately. But my e-book is one file. Why should I manually break it up?! With Inktomi, I could tell the search engine to search only specific tags of the indexed documents. Very Cool. Imagine you are a librarian -- you no longer have to maintain two catalogs -- one for abstracts and another for full text -- maintain just one, but serve both needs.
I know pricing models for software are complex, but Inktomi's metrics wasn't it. Inktomi's pricing model somehow seemed to take into account who owns the content (why should that matter for a search engine pricing?), and number of documents (again, I was not asking aboout a hosted solution here; my hardware, my resources -- why should the number of documents indexed or
search traffic be an issue?). It's like saying "the price of this apple is a dollar, but if you are going to make a pie out of it, it will be $10".
The Last Word
Get Inktomi Search (No, I do not own IMKT stock); your users will be happy.
|First Written: Thursday, August 16, 2001|
Last Modified: 9/8/2001