How would you build a search engine for large volumes of XML data

How would you build a search engine for large volumes of XML data?

Answer:

The way candidates answer this question may provide insight into their view of XML
data. For those who view XML primarily as a way to denote structure for text files, a
common answer is to build a full-text search and handle the data similarly to the way
Internet portals handle HTML pages. Others consider XML as a standard way of
transferring structured data between disparate systems. These candidates often describe
some scheme of importing XML into a relational or object database and relying on the
database's engine for searching. Lastly, candidates that have worked with vendors
specializing in this area often say that the best way the handle this situation is to use a
third party software package optimized for XML data.