stéphane ducasse wrote:
>>The update modifies only one method so it's quite simple to revert that
>>back
>>to the previous version and things work as they should again.
>>Karl
> 
> 

> Yes karl this is true.
> but this means creating a cs, fixing its name, connecting to the ftp, 
> editing the list, uploading....
> while Sam can fix its image in one revert click.

I think this is a question in principle: if many people are using the
update stream, then there a many reverts of them in this case. And they
have to know in advance, *what* to revert. And this is just the most
easy case.

For me the question is, should I update to 3.8alpha unstable or just
3.8alpha for productive work? To leave at the bleeding edge and help a
little with implicit testing.

If there would be an error in the unstable update stream from time to
time, which will be reverted after a small time, then this wouldn't be a
problem. But to work with the unstable stream, and always to be forced
- to follow the ML very detailed, to know, *when* I have to
(incrementally) generate a new image;
- to generate a new image for each buggy update (which is very slow at
the moment from 3.8alpha to 3.8alpha unstable) is frightend for me.

I think there is a tradeoff:
- many unstable testers with more work for the unstable maintainers, or
- less unstable testers with less work for the unstable maintainers.


Greetings
Stephan


> So if someone provide a cs I can do it but I'm terribly busy now and 
> dead tired.
> 
> Stef
> 
> 
> 

-- 
Stephan Rudlof (sr at evolgo.de)
   "Genius doesn't work on an assembly line basis.
    You can't simply say, 'Today I will be brilliant.'"
    -- Kirk, "The Ultimate Computer", stardate 4731.3