
Ariel Oleksiak wrote:
First, it would be nice to know when do you plan to put an updated version of the schema to the webpage. Other, more specific questions are as follows: 1. We noticed that the timeUnits type is not used in the CPUTime element. Is it intended or this is an oversight? 2. The "minutes" unit is skipped in the timeUnits type. I know that it can be easily transformed from other units but this is rather a common unit and the only lacking one in this type.
These two issues appear to be oversights. Thanks for spotting them.
3. What is the difference between job annotation and description in the job identification element?
Description is intended for human consumption, annotation is intended for machine consumption. (I think this is right?) I suppose it is theoretically possible to merge the two (such as has been done in other standards, e.g. WSDL) but the human-readable description bits are much more widely applicable so we're not doing that (well, not right now).
4. What is the difference between cputime in the limits and resource elements? Does it define a limit that must be feasible at a given resource (e.g. by a queue configuration) in the resource element, and an actual limit set for a job in the limits element?
The resource defines how much processing time a job wants to reserve and need not be implemented with limits (e.g. could use a watchdog process instead) and the limit element specifies a POSIX (or semantically equivalent) soft limit on the job (i.e. the job gets a catchable signal when the limit is hit and can therefore in principle have some extra time after that to clean up for e.g. checkpoint-and-migrate). In other words, a job can exceed its limit for the purpose of finishing off neatly, but it cannot expect to exceed its resources in any meaningful way (or the cost for doing it might be prohibitive; that'd depend on the resource-provider site's policy). Does that make it clear? (And yes, I know that the current docs don't explain limits well enough; fixing this is on my to-do list. ;^) Donal.