I agree that the ASP solution is the future, but many of the big companies that we talk to are far from there.
We host a majority of our clients, which are medium to small(500-100 users), but the larger companies all want it inside their firewall.
That's not consistent with my experience. I'm going to meet with a client tomorrow that has roughly 7,000 users and hosts with an ASP. Now, that's certainly not to say that all or even most of the largest organizations are looking for an ASP model. However, by and large, training departments starting to discover what a PITA it is to have to depend on IT to keep their apps running.
Do any of your corporate clients take advantage of the fact that the code is open? The reason I ask is that I don't believe that any of the large companies with whom we are working or speaking that want to host internally would customize our product regardless of what language it is written; therefore, I am not sure that the people with whom I talk care if it is written in TCL. Do you have contrary experience?
First, to answer you question, none of my clients take advantage of the open code because none of them are using dotLRN yet. There isn't even an official dotLRN release yet AFAIK, so it's been just about impossible to market. That's all changing now, though. But in terms of customization, it's tricky. It's absolutely true that many training departments don't think they would be interested in customization per se, and the biggest companies have access to the internals of even closed source programs. So ease of customization is probably not a good selling point for this market (yet). The better argument is that dotLRN is built on a far better technology platform than the competitors. Nearly all current online learning server-side programs are basically unusable. Part of that is because we're still learning how to do this stuff in a way that's useful, but much of it is also because these apps are designed by bad programmers working in capital-starved closed source companies and supervised by marketing drones. The stuff is just plain bad. Heck, WebCT wasn't even backed by a real database until very recently. I know folks who have to restart Saba practically on a weekly basis.
Beyond that, the appeal of Open Source here is that e-learning apps are not close to being mature yet and the closed source world has shown no sign that they can innovate in this area. Even if you aren't going to customize, the fact that others are is a huge win. I read somewhere recently that, on any given Open Source project, the number of people who submit bugs far outweighs the number of people actually writing significant amounts of code. These organizations can contribute simply by banging on the code and telling the community what breaks. And they'll get far more satisfying response time for fixing the app than they will with some shrink-wrapped product made by a vendor that barely makes enough money to keep it's doors open.
Also, when I am speaking to a purchaser in a fortune 500 manufacturer, all he cares about is that our system addresses a pain, that it is going to save the company money, and that his IT people don't have a problem with the technology. That is where we run into the AOLServer dilemma. Although I am a strong believer in AOLServer, I have unfortunately been unable to overcome this dilemma in many cases. If anyone has been consistently successful selling AOLServer, I would be very eager to understand the sales approach.
Like you, I also believe that AOLServer is possibly the biggest barrier to wider adoption. Up until now, that hasn't been a problem because the natural market for OpenACS has historically been hosted apps. But the market is changing. Still, my perception is that the community has some time to wrestle with this issue and come up with the best solution, whether that's better packaging and marketing or an eventual migration.
If a funded effort was devoted to porting mod_aolserver to Apache 2.x, do you think that we could ever get performance comparable to the AOLServer configuration?
The truth is that nobody knows. Several of the uber-hackers in the community are on record as saying that they think Apache would be slower, but I'm not sure if anybody has a deep enough working knowledge of the Apache 2 internals to know just how much of a difference there would be without investing some time in tests and experimentation.
How much funding do you believe it would take?
Personally, I have no idea. I would think that the first step would be to fund a partial port that would (a) give us a better sense of what we'd be up against and (b) give us some benchmarks on representative functions that would give us a clue as to eventual performance.
Does the Apache community have solutions like the OpenACS and dotLRN?
No, not that I know of.