deletions | additions       diff --git a/sectionDISCUSSION_We.tex b/sectionDISCUSSION_We.tex  index 2dc904c..fdd7b35 100644  --- a/sectionDISCUSSION_We.tex  +++ b/sectionDISCUSSION_We.tex 
...
\subsection{Discussion of Scheduling and Failure Data}  A clear and dependable accounting of the user operation time is a prerequisite for a meaningful operation reliability evaluation of an accelerator.   To draw an extreme picture: if all beam outages are just declared to be shutdowns in retrospective  then the beam is always 100\% available during user operation. Which of course does not say anything about the reliability of the accelerator.  We propose to distinguish three accounting of user time:  the originally scheduled time called "Scheduled User Experiment Time";  the planned time slots to re-schedule users that could not finish their experiments, called "Scheduled User Reserve Time";  and the time given spontaneously to users, for example a cancelled machine shift that is converted to user operation just a few days before the actual shifts, the so called "Spontaneous User Time".  We want to distinguish between shifts given to the users at short notice and extra shifts announced to the users at least a month before.   While it is a common practice to compensate users for day long outages by extending the user run, these extra shifts can often not be used by the users: other liabilities often require the users to leave and the extra time is sometimes only used by half of the beamlines.  If the additional time is announced months in advance then this allows for some flexibility for the beamline responsible to reschedule their users; leading to a much better utilisation of the extra time.  But in reality most facilities do not have Scheduled User Reserve Time.   And some facilities do not need it: if the users are predominantly local, then they are more flexible to adapt to changes in the schedule on a short notice.  If a facility is operated on a five day per week basis, then the Saturdays are often used to extend the user runs and the users already plan accordingly.  Some facilities do not provide any extra time for the users: if the beam was down, the users have to re-apply for beam time in the next proposal period.  Therefore one could consider to drop the distinction between Scheduled User Reserve and Spontaneous User Time:  it would not change much for the common operation metrics, the data would still allow to assess the reliability of the accelerators.  But it It  is as well  very important that only those time times  can be counted as User Time where all types of the  failures are recorded. If failure modes within the extra time for the users are not analysed, then this time must not be accounted as User Time in the common operation metrics data. 
...
But the more failure modes are analysed, and the more operation modes a facility runs in,  the more work it is to collect, categorise and evaluate the failure mode data.  We therefore recommend an automated, rule based failure recording into a database, that allows to assign systems and root causes to each failure and enables an automated evaluation of the failure data.