I would play on:

• licensing costs (WebSphere is expensive)
• training costs (WebSphere with VisualAge for Java is a sure productivity killer - just been there)
• rapid development and incremental approach (Perl) vs. framework/waterfall/customization approach (WebSphere)
• interfaces to external systems - Perl is adaptable, WebSphere either provides adapters or you are lost
• complexity - is something like WebSphere really needed
• how big will the development team be - with Perl you can get something done with a few people (one to five), with WebSphere you will have to have many different specialists (three upwards) to get anything done
• show the client a convincing solution for support, with respectable names on this
• provide respectable references that use Perl for commercial applications, e.g., from the DBI references list

I think you will lose, though. I have tried to argue this exact case with several big clients, and lost all arguments (might be my fault, of course). Try to define a cut-off point to decide how much resources you will spend on trying to convince the client. Otherwise this might become an acquisition resources black hole. (Extremely sorry to say this, but it is my experience.)

Christian Lemburg
Brainbench MVP for Perl
http://www.brainbench.com