Help Interpreting traceroute output

[trlists at clayst.com (trlists at clayst.com)]

I am troubleshooting a performance problem for a client that involves 
installing a software package from a mirror site.  The folks who built 
the package, and the installer, think it's our server or network that 
is slowing things down, but we've convinced outselves it's not, and 
we're looking at wider network throughput issues to see what that 
shows.

In doing so I've been using traceroute on several different Linux boxes 
and the results don't make sense to me.

Before I get into the details, let me check something basic.  In the 
traceroute output, is the time shown the total time to the listed 
gateway, or the time from the previous gateway to the listed gateway.  
For example (just showing one of the three trials here), in this 
output:

 5  oc12.Level3.nycmny01.us.bb.verio.net (209.244.160.150)  204.218 ms  
 6  p16-1-1-0.r20.nycmny01.us.bb.verio.net (129.250.2.36)  188.715 ms  

do I read it as calculated values of 204 ms from gateway 4 to 5 and an 
additional 189 from gateway 5 to 6, so that the times add?  Or as raw 
values showing that a test packet made it to gateway 5 with a TTL of 
204 ms and the next one made it to gateway 6 with a TTL of 189 ms 
(presumably because things were a little faster when the second packet 
was sent)?

The latter is the only way that makes sense to me, but several people 
who have been around Linux a while have claimed that the times are hop 
to hop, not cumulative?

Thanks,

--
Tom