This is a continuation of bug 154, "slow responses", because we've made some progress through the woods and there's a bunch of writing there that we don't need to slog through any more.
There is an ongoing problem on the lalashan/yushan wikis where the end user's browser sends an HTTP request in a TCP packet, the server doesn't receive it, the browser sends a series of TCP retransmit packets, the server responds to all of them, the browser (I think) doesn't receive any of those responses, and nothing happens from the user's point of view until exactly 5 minutes later, when the browser resends the original request and the server's httpd responds by finally invoking the wiki software to handle the request.
We have established with authority that this is not the fault of the MediaWiki or WorkingWiki software, because the problem is that they are not invoked until 5 minutes too late. We have also established that it's not just a problem with the connection between my site in California and the server in Ontario, because we recorded it happening to Jonathan while he was at home in Hamilton.
I suspect it's happening to our wiki users a lot, and they don't have a way of telling whether it's a problem like this or the wiki software is slow and badly written. This is a problem for WorkingWiki, because lalashan is our flagship site, where almost all our users are, and it's frustrating them and making it hard to use the wikis. This has been going on for more than a year.
We're considering the possibility that the issue is due to bad transmission of packets somewhere outside of lalashan. If so, what are we going to do about it? It might fall outside of RHPCS's domain, and it's clearly outside mine. But if we keep at it, I'm sure we can eventually pinpoint the problem.
Anonymous