One option smart needs to conquer the world

[Jeff Johnson]

On Feb 22, 2006, at 12:40 PM, Neal Becker wrote:

> On Wednesday 22 February 2006 12:09 pm, Jeff Johnson wrote:
>> On Feb 22, 2006, at 11:02 AM, Neal Becker wrote:
>>> When I talk about smart on fedora lists, there is a lot of
>>> resistance to the
>>> idea of a package system that would automatically downgrade
>>> packages.  This,
>>> they say, is not safe.
>>>
>>> What about a configuration option to prevent downgrading, or only
>>> allow it by
>>> manual override?  That should satisfy everyone.
>>
>> Hmmm, there's actually a better implementation of downgrades  
>> possible.
>>
>> What stresses users with downgrades is that vital or important
>> information
>> might be lost forever.
>>
>> Repackaging the files that were present and saving makes downgrades
>> less risky.
>>
>
> I don't think that's the objection.  The objections I've heard  
> stated are that
> downgrades may be unsafe.  It is argued that upgrade paths are  
> tested but
> downgrades are not.  Personally I have no opinion on this.


"... may be unsafe ..." is more Fedora FUD from the cluelesser lusers.

There is a symmetry principle that helps guarantee that a rollback  
(which is not
a general downgrade, the elements were known to upgrade) is safe.

But yes, if your packages are crap, and you can't/won't do the QA  
necessary
to insure that the package that is being rolled back, indeed, has no  
flaws in
erase scriptlets (i.e. these are the only scriptlets that are not  
already known to
"work" by the reversal symmetry I mention), well yeah, rollbacks  
won't save you.

"Unsafe" my ass.

73 de Jeff