This is one point that my colleagues were complaining about.
We have strictly locked down computers and installing a client on them is not an option in most cases.
Submitting each file separately is quite labor intensive and discourages the review.

Is anybody actually reading this?

This should be a manual override for filename-matching or extension-matching.
Prediction of file name format is impossible for all cases.

jasoncohen: No, syntax highlighting has to be selectable, because the name and extension may not tell anything useful for deciding how the contents should be interpreted. It is along the lines of "html in browser view (for design reviews)". Different ways of highlighting/showing for same file type may be fit for diffrent times.
Selection override can be hidden away.

* Ian T. Johns: I acknowledge that with simple revision changes automatic detection may be good enough. What I'm worried about are border cases. Like accidental commit of a version that has commented code removed ... followed by a fixed code ... just to make it more "soap opera" story the new version also fixes formatting defects that were on each line.

I want a tool that can help with moving to better code and can start from "a horror story" legacy code. The tool should help the transition, not impose more assumptions on the code/development.

*Tiffany Cooper: yes, it should allow multiple lines and multiple instances.
Example: defect #1 for lines 2,4,6 (1a); same defect (1b) on 5,9,13 and (1c) 7,21,23.

Let me repeat: manual adjustment of line pointers for different version is a must for this feature to be fully usable.

Garen: there should be a distinguishing between those two uses.
Multiple instances of the defect should count into number of defects. Lines involved in a single defect should not. Example: buffer overflow - you should mark declaration and wrong use. There can be multiple buffers and usage places that exhibit same problem.

I have heard the counter argument about finding corresponding lines in later revisions of the file. I think that is a weak argument.
Manual override of the of lines affected for given version would take care of that.