[trustable-software] Trustable Systems:A Chain of Custody

[John Lewis]

In response to Edmund’s comments

___________

I can agree that in a purely Agile world defining tests/acceptance criteria is sufficient but the main restriction on Agile is that in the real world programmes cannot be entirely Agile - different approaches are needed.

Agile is still essentially “programming in the small”. DWP claims that Universal Credit is the largest Agile programme in the world but it isn’t all Agile and in any case, in terms of software development it is quite simple/small and has not yet been delivered.

Added to that, most organisations are fairly new to Agile and the technologies needed to support it (CI/CD - DevOps) are still maturing. As you rightly highlight it all comes down to the CMDB - which is hard to get right and maintain across all the environments that are needed (dev, test, ptp, prod, fo).

Given this lack of maturity, Agile development/Cloud deployment is low CMMI (1-2). And for Trustable Systems, a Level 5 is required. I remember going to NASA/IBM Houston in 1990 (the first CMMI 5) and being shown the paper-based Chain of Custody system they had. The audit team was bigger than the Dev team and that raises the issues of checking the CoC (an automated tool?) and what happens when it is broken?

We agree that Trustable Systems are desirable and necessary for the deployment of secure/safety/mission critical systems, especially IoT and that we should continue to strive to achieve them, but I feel that we still have a very long way to go. The fact that horrendous security exploits are still being found in Open Systems makes me even more nervous of trustable IoT deployment.

regards