Disinfotainment: "Now Dave's solution is to release a script that continuously polls Radio and restarts it if it's crashed. Wouldn't it be better to fix the bugs that make it crash in the first place? Instead, we have a workaround that is even more of a CPU burden."

First, Dave or UserLand did not release the script, I did, so I'm the one to blame.

Second, in yesterday's post I did say that this is not an elegant nor an efficient solution, but that it kept a server running for 34 days. Of course, this is not the best way to run a service and, btw, Radio is also not the right application to run such a service, while Frontier is. We have Frontier servers which have been running for months, serving millions of pages, without much problems, but it took a lot of time and hard work to properly tune them and get them to run smoothly, just like with any other server software.

I don't think that using Radio as a server is a clever idea, unless we are talking about very small communities of users (such as small intranets) where RCS can perfom decently. In this case a low cost computer, an incredibly low cost application and a simple hack, in our case, did the trick.