DSO / logic analyzer buying advice

I currently have a dual-channel 30mhz analog scope, the Intronix 34-channel 500mhz 2k samples max logic analyzer, a Saleae Logic 8-channel 24mhz unlimited samples and a (much) older 60mhz DSO on it's last legs. The 30 & 60 mhz analog scopes are simply too big and cumbersome to use. These new scopes are much smaller and more convenient. For logic analyzers, I have FAST, WIDE, but short buffer. And a slow, narrow, mega-long buffer(limited by PC RAM more or less). These both have their good points but to round out my tools, I need something that ballparks 200mhz, 32-channels, and decent buffer. What's a decent buffer? I don't know exactly. I know 2k isn't even close. And that a million+ is really nice. These are both PC-based logic analyzers and I'm fine with that. The software is sufficient with both and gets me the measurements I need. Of course, I'd prefer standalone units but still would like to export data to a PC (say via USB or serial). This would simplify the process and the UIs of dedicated devices have usually been thought out a little more than the PC software..... I was seriously looking at a Rigol DS1102D which is 100mhz, 16-channel LA + 2 analog channels. My only real beef is that it's 16-channels and not 32. Could really use 32 to do a 16-bit data bus and other stuff (partial address, etc) I can't easily tell looking at the specs what the depth of the LA would be. Would I simply divide the 1 million point memory by the 16 channels? http://www.rigolna.com/products/digital-oscilloscopes/ds1000d/ds1102d/ Rigol's have gotten great reviews. Or do I get a scope only, and then use something cheap like this http://dangerousprototypes.com/docs/Open_Bench_Logic_Sniffer and solder another 16-pin header onto that and go 32-channels, 200mhz, 6k depth. Or buy a kickass eval board and hack SUMP onto it, and get a deeper solution..... While I'm a huge fan of HP and Tektronix, and I know there is old hardware at reasonable prices --- I don't want to mess with trying to piece parts a large mainframe together from untrusted sources, and spend time trying to fix something like that. Plus, the specs are normally not even close to that of modern equipment. I'd rather invest time hacking verilog onto a new eval board that has plenty of fast memory............... Any discussion would be helpful. Thanks, Keith