Attendees

• Alan Froggatt

• Andrew Carr

• Colin Eberhardt

• Matt Richards

• Antony Welsh

Minutes

We debated the question “Do we view the Generator as a financial services specific tool, or as a more generic utility?”

• General consensus was that this tool is potentially useful across a very large number of domains, we should serve financial services, but also as many other domains as possible

We debated “The Generator has at times been considered a testing tool, a data simulation tool, a load testing tool - which is it? Should it be all of these things?”

• We discussed creation of violating data, and it was considered to be a useful feature, but the question asked was, “what is the relative cost of supporting this feature?”

  • It was noted that violating data could be described by the profile without the need for a ‘violating’ mode

• We determined that the cost of implementing violating / testing is high

• General consensus was that this tool should generate realistic simulation data in the short-term

We debated how we prioritise feature development, should we tackle FS features first?

• We have seen a number of non-FS teams wanting to use it already

• Non-FS domains tend to be more simple, making them a good starting point

• There was a general feeling that relational data is a priority

  • however, there are still a number of use-cases where it is useable without relational

  • Without relational, the tool is better suited for creating CSV files, API outputs, but not the population of database tables

• Consensus was that relational data is a v2 priority

• We briefly discussed how this tool should be ‘packaged’

  • We have found that people were wanting to use this tool in various different contexts, streaming direct to database, used within Python, etc …

  • Our current CLI approach doesn’t close any doors yet

We debated “Do we need to invest more in the Generator UX?”

• We agreed that the user experience is key and a high priority

• There was a brief chat about the value of allowing people to share profiles or components of profiles

• We debated whether the generator (CLI and profile) is sufficient

  • There was general consensus that this is sufficient for v1.0

• We discussed that the DataHelix team should spend some time on the profile design, using the steering group to discuss options.