Durug,
A DSNless connection stored in the registry? Well... the performance difference between a dsn and a dsnless connection is the registry lookup for a dsn connection... so... I guess I miss your point. What would be the benefit, and to be honest, I've never heard of storing a dsn-less connection in the registry anyway. Any more info on that?
And as far as application deployment goes, DSNs will make your life easier in that respect... not dsn-less.
Ovatvvon,
Yes, I was referring to that experience several months back. Basically, if I used dsn-less, then the server would have to go out and do a DNS (not DSN) lookup to find the server that was right above him. That lookup took place over a 10mbps line, and then communication between the two machines would also take place over that same line... On the other hand, DSN would let them communicate on the same subnet over 100Mbps (which has since been upgraded to fibre, so it's at the speed of light now ;-) -- not really that fast, but I just like saying that because it sounds cool). So I realized a noticeable difference in my application performance when I changed the connections to DSN.
In retrospect, I suppose I could have used the subnet IP address to accomplish the same thing. I didn't think of that at the time, but as it turns out, because of the portability issue I described above, I am very happy with my decision.
Now, as far as high traffic (and I define that as at least hundreds of concurrent connections) goes, I couldn't tell you honestly. We do many different things w/ our servers... all having to do with research. So we do reporting, surveying, ordering, etc...
On our high traffic days, I havent taken the time to go in and see what's taking the resources. The surveys use a separate process for each user... third party solution. While all reporting and such are custom written db applications, which use connections to our SQL Server. We have good days (response wise), and mediocre days... depending on what we have in the field, and like I said, I just haven't bothered to measure exactly what's eating the resources on the bad days. My suspicion is the survey software (which has nothing to do with db conns), but that's just a suspicion (and a likely one at that, seeing as I write the db stuff ;-))
You know what I might do to try it out? The next time you write an application that you expect high traffic on, I'd code it such that the conn is stored in and asp include, and first make it dsn-less. Then, wait for traffic to go up... jump on there and get out your stopwatch and run some routines. Time em.
Then, (very quietly) sneak in there and change the connection to use a DSN, and then get your stopwatch back out and go run the same routines. I would really be surprised if you noticed any difference, but I might be completely wrong on that one.
If you do that, though, please let us know what you find. I might even run a little internal test this week since now you have my curiosity peaked. If I do, I'll also let you know what I find.

paul