I want to be able to use TERM() on complicated values such as HTTP URLs and user agent strings. Usually, this does not work due to spaces and other breaking characters.
Is there a way to customize the terms at all? Hopefully on a sourctype and field basis?
If not possible, would other solutions like summary indexing help>
Altering what segments can be searched with TERM involves changing the segmenters.conf: http://docs.splunk.com/Documentation/Splunk/6.6.2/Admin/Segmentersconf
Be aware that changes in the characters for major and minor segmenters in the default stanza will have a big effect on ALL data indexed and the disk space required for indexed data. To affect a particular sourcetype, reference it in the stanza name.
.. Or to your other question and more simply/safely, you could just create a transforms that indexes a field of interest at index time which could key on an indexed field and would be called from props on a sourcetype basis.