Bug#27906: PROPOSED] Binary-only NMU's

[Paul Slootman <paul@wau.mis.ah.nl>]

On Mon 19 Oct 1998, Ian Jackson wrote:

> I wrote:
> > [...] there's no harm in a small amount of version skew at release time.
> 
> Several people have misunderstood this; my apologies for being
> unclear.
> 
> I meant that there is no harm if the binary versions for (say) m68k
> and i386 are slightly out of step.  So, there's no need to rebuild
> i386 binaries just because the m68k porter did an NMU.

OK, so the i386 binaries are permitted to exist without a corresponding
source version (as the m68k porter was obligated to do a sourceful NMU).

> I did NOT mean that there is no harm if there were binary versions
> without corresponding source.  That's the opposite of what I mean !

So you were contradicting yourself; because you didn't state that:

> Since we cannot rebuild for all architectures simultaneously and do
> not want to withdraw binaries or wait with porting,
> *we MUST be able to have more than one source version in our archive*.

It indeed boils down to the necessity of being able to have more than
one source version in the archive at the same time. You didn't state
this clearly at the time where you talked about "there's no harm in a
small amount of version skew at release time".

I'm still not clear on what you mean with "at release time", however.
Surely all of this is applicable at _all_ times? It's not like we
porters only start doing stuff at release time; it's a continual
process. Also, the archives are accessable (and mirrored) at all times.

> look like ?  So far we have seen two proposals:
>   i.  Simply have them side by side, with some kind of way of making
>       obsolete sources disappear eventually

This one is it.


Paul Slootman