Problem when graphing statements using timestamp / stored

[davetosh]

At the moment we are using stored to graph statements. Clearly this is not ideal as graphing should really represent the actual time the learner undertook the particular action and the statement was generated. However, with timestamp only being optional - this is not reliable.

I would love to see timestamp become required in the spec. What do others think?

[fugu13]

There's language in the spec about this, in the timestamp attribute: "If not provided, LRS should set this to the value of "stored" time."

It at the very least gives you the option to do so (and I highly recommend it), though IMO it should be updated to a MUST; the language predates clarity over the should/must distinction, and there's no good reason not to require it of an LRS.

[ht2]

I'd concur - seems like it should be a required field, especially if we move to a system of queuing large influxes of incoming statements. In those cases the stored timestamp wouldn't accurately represent the issued time. Another potential issue would be if a large array of statements was stored without a timestamp, the analytics would show a significant spike during that time!

For the moment it isn't a problem - at very most the timestamp will only be a second or two off the actual value.

[fugu13]

Since you can set stored to moment of intake, that isn't a problem, though if people are sending large volumes of statements that is a problem -- but on their end, not the LRS's end.