Providing better "recovery" within official scripts

[Mikus Grinbergs]

This is probably the wrong list - but I would like developers to be 
aware of this:

Just installed hardy, using 'update-manager -c'. The script abended. 
[I had last year seen 'update-manager' fail trying to install gutsy, 
and earlier yet had seen 'update-manager' fail trying to install 
feisty.]  The current failure with hardy was because of not enough 
room on /boot - but I was able to ipl the new hardy version anyway.


My complaint is that the only "recovery" which 'update-manager' 
appears to provide is to __undo__ everything.  Having spent x hours 
downloading the new packages, "back to zero" was not acceptable.

"Official" scripts (such as 'update-manager'), when they encounter a 
problem, should not present the user with a single option ('undo'). 
They should allow the user a second option -- "keep what was done 
already" -- and put in some intelligence, such that if (re)starting 
the script finds a partial result "kept", the script will bypass 
performing those steps which were done already (and "kept"), and 
will resume with the step that failed before.


mikus