Re: [boost] [epochs] Proposal for an epoch-based organization of Boost libraries

On May 27, 2020, at 4:57 AM, degski via Boost wrote:

Although I think the mechanics of the proposal are well worked out as I posted earlier, but on reflection, I don't think this is the right way forward. The world (linux had already) is moving to live-at-head, see the development of vcpkg. I consider this an old-school solution to a new-school problem, so it will stumble on. I think, that IIRC what pdimov5 suggested somewhere last year, to re-baptise master -> release, develop -> master, experimetal -> develop, and just live at head, is the best way forward. Doing a release comes down to adding a tag to a master-commit and everybody carries on busying themselves in experimental in the meanwhile.

Just to be clear, do I understand you to mean by “live-at-head” that formal releases, with appropriate quality control, etc., are not a desirable requirement? It seems to me that there is indeed a trend in that direction. Increasingly, software seems to live in a repository somewhere with no releases whatsoever or very few at intermittent intervals. This means that quality control decisions are entirely left up to end users, who of course are rarely knowledgeable of the internals that in fact determine quality. At the very least, that places a large burden (researching each project and learning all the details) on a large population (the many users), to avoid a task (identifying a quality release) by those best positioned with the relevant knowledge (the author(s)). Exactly the opposite of the goal of reuse. I hope this is not your suggestion and that I have misunderstood. However, the clear trend elsewhere in this direction makes me wonder, so I would appreciate clarity on what you really think the “right way forward” is. Sorry if I mistook your intention. Thanks. Chees, Brook