Re: booststrapping /usr-merged systems
On Sun, 11 Jun 2023 at 22:17, Timo Röhling <roehling@debian.org> wrote:
>
> * Luca Boccassi <bluca@debian.org> [2023-06-10 19:54]:
> >I would caution to avoid interpreting clarifying questions being asked
> >as dissent. It's good to ask questions and clarify details about
> >corner cases, but I wouldn't automatically write them down as
> >disagreement. At least that's my reading of recent parts of this
> >thread.
>
> This is also my understanding. And for the record, I want to
> emphasize that I am very much in favor of the plan that Helmut came
> up with, for a number of reasons:
>
> [Full disclosure: I had a few in-person discussions with Helmut in
> Hamburg last month, so I am probably somewhat biased by now.]
>
>
> 1. Helmut has shown experimentally that his transition plan can
> work. There are always unknown unknowns, of course, but at the very
> least, we do not have to break any use-cases intentionally.
>
> 2. The transition will leave us in a well-defined state post-trixie
> without the need to add (and continue to maintain) any clutches
> (or "special cases") for dpkg.
>
> 3. Almost all problematic cases can be dealt with by some black
> magic in a single usrmerge-support package. It is not pretty, but it
> will get the job done; a bunch of trickery to make dpkg do the Right
> Thing despite its incomplete knowledge of aliased paths.
>
> 4. We will be able detect the few cases where the Right Thing does
> not happen transparently, and we can even give advance warning to
> affected package maintainers what they should and should not do. If
> the maintainers of those packages pre-upload their transitioned
> packages to experimental for some automated tests and verification,
> we can avoid any breakage in unstable and testing.
>
>
> Of course, you do not have to take my word for any of this. I am a
> big fan of Helmut's approach with experimental verification and
> data-driven discovery. Have a look at his published test scripts and
> try to poke holes in them. The more people do this, the more
> confidence we can have that this might actually work after all.
Hi Helmut,
Any update on this topic? I believe you were working on a write-up,
how's that going?
Kind regards,
Luca Boccassi
Reply to: