Suddenly the remarkable growth ceased and there was a marked resurgence in the usage MSIE.
To my surprise I discovered that MSIE accounted for 70% of browsers visiting this site. Mozilla (including Firefox) has slumped to 18%, and Netscape and Opera were still clinging desperately to their small margins, each struggling desperately to keep their heads above the one percent threshold, as they have been for many months.
Most dramatic however, was the turnaround in OS stats. Various forms of Microsoft OS account for 90% of the visitors to the PGTS site, while Linux seemed to have slumped to a record low of 3%. Mac OS X was still a healthy 4% and looked steady.
Although the figures for the PGTS site show a bias, based on the type of visitors who come here, I have gone to considerable lengths to make them as accurate as possible. This is not always the case with the surveys of agent strings (due to the complex nature of agent strings and the lack of widely-accepted naming conventions)
I checked other sources, these trends seemed to have been confirmed by figures from elsewhere. But they were nowhere near as dramatic as the figures I was seeing. Most of these trends move in at a snail like pace (except FireFox which until this month really did seem to be on fire!)
Then I started to contemplate the trend ... it was so dramatic that I began to suspect that something might be amiss.
Then I looked at the figures for robots and I realised that something was definitely wrong!
GoogleBot was sitting in Fourth position! Then I knew that there was a problem with my update script (a perl/postgres script that updates the PGTS agent_string database).
I checked the log file for the update script. Sure enough it was crashing halfway through the run. All the more recent agent_ids (like FireFox and the new GoogleBot) were not being updated. It appears there may some junk information in the Apache log file which causes this.
I will have to fix this on the weekend!