Comment 11 for bug 1777675

Mike Rylander (mrylander) wrote :

Thanks, Elaine!

WRT inventory history, one problem with looking at the historical scan date alone, including just the "last" inventory time, is that without some sort of session and physical locality concepts to provide grouping, it can quickly become very difficult to identify (and even to analyze, short of some time-series visualization where a human looks at the peaks on a graph) when scans are nominally "concurrent" for an inventory process.

It's certainly useful to see that a copy has not been inventory-scanned within the last year and its status is Available, but things become murky when you want to see the state of the collection in real-time. If the scanning process, for a particularly large or particularly understaffed library, or even just particularly unlucky library, takes a while to complete -- perhaps done over the course of a few months -- then during that time you can't really use "last scan time" alone reliably to indicate presence or lack thereof.

The fact that those situations can't be reasonably covered by the simpler proposed feature does not mean I'm against the simpler feature existing, it's that I don't want the simpler feature to make the fuller feature significantly harder to engineer down the road by having to take the existence of fields (for reporting, etc) into account in a future implementation. If we make a reasonable guess about how just the data we need now would fit into a more full implementation, we can make our future selves happier.

(I can reiterate my database-level concerns more if that's unclear, but they require some preexisting understanding of how PG works internally. This bug probably isn't the forum for a PG-internals workshop -- I've provided tangents enough thus far. ;) )

[As a side note, with historical scans you can, at least to an approximation, investigate the collection by each run's bounding dates to see changes over time, such as identifying sub-collections with more "stock lossage" as a percentage of the part or the whole. That may not be important for annual reporting, but it seems it could be important for placement and security.

Of course, with the appropriate run and collection grouping mechanisms of the full proposal (granted, the cost is non-trivial) those sorts of reports and analyses are both natural and well-defined.]