Definition of a "blocker" bug ?

[Christian PERRIER]

Quoting Jeremy Allison (jra at samba.org):
> Volker and I ended up having an interesting
> discussion in a bug report about the meaning
> of the term "blocker" as a severity catagory
> in bug reports.
> 
> I have a definition I'd like to propose -
> 
> A "blocker" bug that will prevent a release
> is one of the following problems:
> 
> Security holes.
> Data corruption.
> Resource leaks (covers fd's, memory etc.)
> Unrecoverable client-visible errors.

Clearly identified regressions, limited to those that are regressions
from a former version in the same series (bug <foo> does not exist in
3.4.2 and appeared in 3.4.3)?

I was about to say "any kind of regression" but that may lead to
difficult cases when a regression is identified wrt a really old
versions's behaviour and may then be related to more complex code and
behavioural changes.