At the moment the OOo crash-reporter enables us to better find bugs inside the
code by sending a stack trace of a recent crash to a designated server. This
server collects and processes all these reports. Although it is possible to
count the total number of reports per OOo release in a given timeframe, it is
still not possible to measure whether the stability of OOo actually improved
from one release to another and how much. The reason is that we do not know
whether e.g. an increase in crash-reports per month for a given release can be
attributed to the release's decreased stability or to an increase in usage of
this release.

One possibility to extend the current data sent with a crash reports is to add
the time which OOo has run until the crash occurred. Even in the current version
of OOo, a log-file entry is written on the start of OOo which stores the time of
the start. When crashing, the crash-reporter could read from this log-file entry
and calculate the runtime until the crash. In case of subsequent starts and
shutdowns of the programs without any crashes, the runtimes would have to be
added. This measurement of the average runtime until a crash occurs could then
give us an impression whether the product quality increases or decreases and it
might also give a rough impression about how much it did increase or decrease.
E.g. an average runtime of 10 hours until a crash is better than an average
runtime of 5 hours until a crash.

Although there are some difficulties with this approach (e.g. the measurements
are influenced by usage of the OOo quickstarter or the user suspending his PC),
it still seems the most promising approach to get at least a rough estimation of
stability changes with different releases of OOo and thus helps us in increasing
the quality of our product. Therefore the crash reporter should be extended to
include the time OOo was running since the last crash (or first start ever) into
every crash report.