Thinking about it some more... what we're actually talking about here is a file
update policy. An update policy is a selection of operations such as whether to
replace modied files or not, to backup or not etc. When looking at it from this
perspective, it all starts making more sense: %config is merely a *name* for a
particular selection of such operations, aka policy, not a feature in itself,
and this new thing is not a parallel feature to %config, but another update
policy which partially overlaps with %config.
So implementation-wise, rpm internals need a bit of generalization (what a
surprise) before proceeding. And of course it needs a name too, but for
starters %updatepolicy(...) will do, better suggestions welcome. After that,
%config can be made into an actual macro that emits a matching %updatepolicy,
and maybe by that time we've come up with a couple of other policies common
enough that they deserve a name in style of %config. %ghost related to this
too, but its roots are deeper since the data is omitted from the payload
entirely.
As for the behavior the actual RFE, I suspect there are cases where a packager
*really* wants just the initial content, never to be touched by rpm again. So
that needs to be an option along the "update until modified" variant. I guess.
Then there's the question of legacy compatibility, all this should be
retrofitted in a manner that makes older rpm versions not implementing the new
thing to do something relatively sane. Adding an rpmlib() dependency on this
would be like killing a fly with a sledgehammer.
--
You are receiving this because you commented.
Reply to this email directly or view it on GitHub:
https://github.com/rpm-software-management/rpm/issues/152#issuecomment-293532010
_______________________________________________
Rpm-maint mailing list
Rpm-maint@lists.rpm.org
http://lists.rpm.org/mailman/listinfo/rpm-maint
