Please add tika libs in WEB_INF lib directory

Nice, though I wonder what the use case is. Do we need this?

Using tika as servlet will allow non java programmer to use tika. I have done this implementation for a team composed from .NET programmers who wants to extract content and metatdata from documents without using Java.
The only think they have to know is an API to process XML in there specific programming language.

I haven't looked at the patch, but I would suggest that it be viewed as a contrib dependency. I think one of the things that will really help adoption is if it is easy to remove any dependencies that aren't needed for a given application. That is, we probably have a core Tika, and then Tika contribs.

Also, note, SOLR-284 is also related.

Another alternative for cross-platform use is the CLI feature:

1. Extracting structured text content from a file
   java -jar tika-0.2-standalone.jar --xml /path/to/file

1. Extracting plain text content from a file
   java -jar tika-0.2-standalone.jar --text /path/to/file

1. Extracting metadata from a file
   java -jar tika-0.2-standalone.jar --metadata /path/to/file

This way you don't need a separate server process and there won't be any concerns about unauthorized users getting access to your files.

I'm a bit concerned about any web service that allows the client to retrieve the contents of any file on the local file system. Would it make more sense to always require the client to upload the files they want parsed?

Also, the file system traversal feature seems a bit outside the scope of Tika, though having something like this in a contrib area might be nice.

Also, the file system traversal feature seems a bit outside the scope of Tika, though having something like this in a contrib area might be nice.

I believe Droids (crawling) has integrated Tika already as well. But, yeah, as optional contribs, those make sense. We will have lots of dependencies on extraction libraries as it is, so I really think it makes sense to stay as lean as possible elsewhere. Before you know it, Tika will be a 50-100 MB download, and that will slow adoption...

I see a servlet making quite some sense - think of Solr, but only having the extraction request handler... That way you could have a central meta data / text extracting server without needing to install java + tika on all the hosts where you might need it in a replicated CMS environment f.e.

So the scenario would be that a CMS trys to extract text, meta data from a file, but does not have a local tika at hand. It would then send the file to a Tika server and get the results back in XML or JSON like Solr does.

Resolving as Won't Fix, since the extractOnly option of Solr's ExtractingRequestHanlder already provides similar functionality.