Finding real-world kernel subsystems

Posted Feb 2, 2021 19:55 UTC (Tue) by nix (subscriber, #2304)
Parent article: Finding real-world kernel subsystems

> Assuming that the process itself makes sense, a project that adheres more closely to its defined process should produce higher-quality software.

Sorry, does this follow at all? It doesn't seem to do so to me. This is only the case if we know that the defined process in question is a local or global maximum in terms of the process producing high-quality software. Just looking at an existing process and formalizing it does not imply in any way that sticking to that formal model will generate higher-quality software, just that it's sticking closely to the model that the software's development process was already following: if that process was bad, following a formal model of it will produce bad software. This seems likely to me to (at best) be a wash in terms of quality, on the average.

to post comments

Finding real-world kernel subsystems

Posted Feb 3, 2021 1:13 UTC (Wed) by interalia (subscriber, #26615) [Link] (1 responses)

> Sorry, does this follow at all? It doesn't seem to do so to me.

Yes I found this a bit puzzling as well. I think the unstated assumption is that the development process is thought to be a good one, therefore following it properly is also a good thing. Given two projects A and B which both follow development process Z, then the quality of A will be "better" than B if A follows the process strictly whereas B only follows it haphazardly. In this situation "better" is not quite true, it's more like the quality of A would be steadier and therefore more reliable than B's quality, which I think intuitively makes sense.

I can see how this might make sense in a certification context where if the dev process is certified but the project doesn't really follow it, then approving the process is useless. It sounds a bit like an audit for ISO 9001 compliance.

Finding real-world kernel subsystems

Posted Feb 3, 2021 11:27 UTC (Wed) by pizza (subscriber, #46) [Link]

As it is rather difficult to objectively measure _quality_, consistency is used instead, with process being the primary way that consistency is maintained. But like all metrics, it ends up becoming what things are optimized for, even at the expense of actual "quality".