But how does the latter scenario differ from one where `mktemp` is invoked without `-u`? Even if the file is created (as opposed to just the name), nothing prevents another process from writing to the same file once it has been created.

Preventing ownership changes is not a filesystem/Access Control Lists feature, it’s a UNIX feature designed to prevent bad things from happening.

I suppose you could give the wordpress user the permission to chown via sudoers if you really wanted to, but I would question the reasons for trying to do this in the first place.

Since you appreciate typo spotting, a couple of more it’ses to itses: ConsoleKit and its successor, to retain its position.

This was very interesting to read, thank you. I wish more people could see these issues as resulting from conflicting design philosophies and would put their effort on improving implementations of their choice, rather than arguing that the other choice is inferior (let alone those ad hominem attacks against the other camp). I for one believe both designs have their uses, and that we shouldn’t be advocating either as the be-all end-all.

From a similar question on ServerFault (and particularly one response there), one possible explanation for the disparity is that there are processes hanging on to files they’ve accessed on /tmp that have since been deleted.

# lsof | grep deleted 

will list such files along with the processes still attached to them.

According to a Xiph mailing list message from 2003,

ogginfo should check most things, at least the CRC integrity of everything.

I don’t know whether ogginfo is available for Windows though. You’d also have to script the recursion part, as ogginfo doesn’t do that (at least on UNIX).