Leave the overall shield drain wire disconnected at one and, and connect it to a good chassis (NOT signal!) ground reference at the other end. I'd initially ground it at the end that drives the most signals into the cable: this is called "telescoping the shield", and effectively extends the chassis of the driver right up to as close as possible to the chassis of the receiver, without contacting it.
Don't chop the free end off right at first, though. You may find it useful to try grounding the other end instead, if you find that you still have RFI/EMI problems. There is really no hard and fast rule for telescoping shields: you do whichever one gives the best noise immunity. What you probably _don't_ want to do is to ground both ends,: connecting chassis grounds together is very likely to create a ground loop, and make problems worse. But be prepared to try it, just to see.
For this sort of overall shield, you want to use chassis ground instad of signal ground so that you can dump as much of the EMI/RFI hash as possible into ground as far away from the sensitive circuitry you are trying to protect: stick it into the chassis rather than Pin 1 of an XLR. Lots of gear is designed with questionable signal/chassis ground relationships, where noise on Pin 1 of an input gets coupled right into the signal path. If you have an overall shield, use it to your advantage to dump the hash far, far away from your preamps and signal receiving circuitry...
This is called the "Pin 1 problem", by the way- there are some really good sites with discussions of the problem, and whole special issue of the AES Journal on it: the June 1995 issue, which was the special edition on shielding and grounding. It's a must have, and is available from the AES for $15 (or $10- if you're a member). get it:
http://www.aes.org. You can see a lot of the material from that issue from Bill Whitlock at Jensen Transformers at this link:
http://www.jeffrowland.com/tectalk6.htm to decide if you want to snag the whole issue...
Hope that helps, anyway.
