I searched jira and your forums and I can't find this reported, so here it goes....
In the last week, two members in our jforum, posted a message and another member's userid was showing as being logged on instead of the originator who was logged on. In other words, under "Logout [Susie]," it would show "Logout [Linda]" and the post would be labelled as being posted by Linda when the actual person that was logged on was Susie! Both members appear to have used the auto-login (when you return) feature. And, they were located 300 miles away from each other.
Have you heard of this before? Any ideas as to what could cause it? [originally posted on jforum.net by vivie]
Rafael Steil wrote:Are your / users behing a proxy? I have seen this problem occur once with an user that was accessing the web through a proxy, which by its turn was doing heavy caching of data.
Yes, everybody is. The problem occurred to two users in one week. It is not a heavy board. It has just 18 registered members.
Rafael Steil wrote:JForum uses some cookies and hashes (each user has an unique hash), so if the proxy "mix" the data for some reason, we may run in trouble.
Cookies are saved locally, so that can't be the issue. The two individuals, as I said, were participating in the class jforum from two different parts of the state of California. How are sessions saved in the database? The "mix" must happen at that level - we need to pinpoint why.
I don't have that extensive experience with JForum as to confirm or deny that, but synchronization issues are not necessary related to "scale", but more with Murphy's law ;).
As an example, 1 month ago we had a similar problem with a webmail package we are using, as a mail a student sent came up on the "sent" folder of a teacher. The reason behind that in the end was that they had logged in exactly in the same millisecond!! The product is closed source so we cannot look at it/fix it, so we are now looking for a replacement.
On a side note, it was mentioned not long ago that JForum synchronization is a bit "lacking" (see thread: JForum - load test), and that's why this possibility came to my mind. [originally posted on jforum.net by GreenEyed]