The tale of ghost: vmware tag

Question

Once upon a time in SF land, there are mass-killing and burnination ritual of vmware tag. This mass-killing initiated at Mar 10, 2011 and lasted until July 6, 2011. The last-killing blow is done by Jeff Atwood himself.

In the present, looks like the vmware still haunted us. You can see it yourself by

• Visit the graveyard old question tagged with vmware-*.

• Click one of them to bring up the question page, for example: How do I get the Activation Key in Windows 2003 Server. You can see there are no vmware tag in the question.

  

• Right click edit button and select Open in new tab.

• Out of nowhere, the vmware shows up in tag list, replace vmware-server tag. The ghost has been revealed himself to us...

  

This happened when you edit other questions as well, for example

• Minimize VMware HD space with multiple snapshots
• Where does ESXi decompress files to during install?
• Kickstarting VMWare ESX 4.1 (Error: No NIC with name bootif)
• VMWare Lab Manager: What's the difference between Capture to Library and Archive to Library?

I tried to compare the Tags data from SEDE. And the result is there are no [vmware] in Tags table. Also, this question - How do I get the Activation Key in Windows 2003 Server - doesn't have [vmware] tag.

So what can we do to kill this ghost once and for all?

Answer 1

It's just about as killed as killed can get. In fact, vmware is in our list of blacklisted inputs right along side amazon. I'm afraid I don't know how those tags are being translated in the edit screen to those banned tagged. This looks like a bug!

Answer 2

After digging out SEDE once more , I found some explanation why this happened. But still I need SE-dev to finding the root cause of this inconsistency.

Inconsistency when viewing and editing post

The source of the problem lies in dual data source when you edit and view the post. When you view the post, web page pull data from rendered content (with HTML form) in Posts table . But when you edit it, web page pull data from raw content (with Markdown form) in PostHistory table. Normally, when you submit the post, both table updated with same data (but different forms). Unfortunately sometimes these two data can be out of sync, like this case.

To prove this, I wrote a query in SEDE to compares actual tags saved in PostTags table and raw tags recorded in PostHistory table. The result almost ~350 questions in Server Fault have this inconsistency. Almost all questions have vmware in its PostHistory entry although it has been converted to another vmware-wildcard tag in PostTags table.

My hypotesis is when Jeff Atwood "manually" converted vmware to another vmware-wildcard, something wrong happened

1. He forgot to update PostHistory table :(
2. He didn't forgot to update PostHistory table but in the next-day one of fellow dev rollback the database :(

As for the solution for this kind of "bug": It can be fixed quickly by SE-developers. Of cource we I can manually edit it, but it'll bump many vmwarequestions in frontpage.

Additional note:

So, this event has happened in the past too. Especially when image links has been converted from i.imgur.com to i.sstatic.net. SE developers has been "forced" to manually baked up fresh post bodies so it matched the revision history version.