rpm 5.x in Th

[Tomasz Pala]

On Sun, Sep 23, 2012 at 14:18:45 -0400, Jeffrey Johnson wrote:

>> i'm sure people want just to get old package back, to revert human mistake of upgrading or some other reason for downgrade, because package is misbehaving, not wanting perfect rollback like filesystem rollback.
> 
> RPM isn't responsible for human mistakes: no implementation can save users from mistakes:
> 	When the data is gone, you lose.
> This applies to erased files, removed packages, and dead disk drives.

The data is not gone - it exists in repackage. The _only_ thing that's
missing is some syntax sugar to ease downgrades.

> Yes. You do realize I designed a
> 	"Transactionally Protected Package Management"
> to handle exactly and only package manager initiated operations?
> 
> There is zero detectable interest several years later, measured by any of discussion or patch
> submission or attempts at using.

What a surprise! - as I already told you, noone expects ACID-style
rollback from package manager, as this idea is broken by design. Rolling
back filesystem requires tool operating on filesystem (not application)
level.

>> call it something else than "rollback", if it hurts your perfect world
> 
> Call its whatever you want, rpm has been able to repackage existing
> content when erasing for most of this century. Users and distros are not
> enabling the functionality, and the RFE's for better continue incessantly.

Apparently you completely don't understand this discussion. In short:
	it is about using these repackages in comfortable way
Nothing more. No undoing triggers. No 'rollback'. Simple downgrade
_package set_ to the state at specified time.

>> i my world, where i deploy software with rpm packages, i do poldek -u package-old-version --downgrade as i do have old versions available in package manager repository. but distro packages are not available that easily, therefore people look into /var/spool/repackage dir
> 
> So implement --rollback in poldek or yum or urpmi or apt or dpkg or smart or zypper
> or BTRFS or even the linux kernel if you wish.

Indeed, this should be implemented in poldek. But rpm itself shouldn't
suggest having function, that doesn't and won't work - so remove this 'rollback'
and don't confuse users.

-- 
Tomasz Pala <gotar at pld-linux.org>