[Newbies] Re: Why do lengthy installations "shut down" Squeak?

[Bert Freudenberg]

On 24.02.2009, at 20:15, Matthew Fulmer wrote:

> On Tue, Feb 24, 2009 at 06:21:10PM +0000, Zulq Alam wrote:
>> Hi Rob,
>>
>> Rob Rothwell wrote:
>>> Just wondering why, since Squeak has processes, when install  
>>> something
>>> large via the Universe Browser, etc... I can't do anything else in
>>> Squeak while it is installing?  Is this by design?
>>>
>>> It seems like it could just be a low priority background task and  
>>> you
>>> could keep working...
>>
> I'm a Monticello developer. I think the lack of loading in the
> background is mostly a historical accident, but it does prevent
> some (but not all) problems where something gets modified
> concurrently with the load.
>
> Monticello 1.6 features true atomic loading, and thus could
> support background compilation of the package while you continue
> to work. It does the compilation in a sandbox separate from the
> actual system, then merges it in using a single primitive
> operation. For more details see
>
> http://installer.pbwiki.com/SystemEditor
>
> I haven't really thought about what would need to change in order
> to support background loading; the biggest change would probably
> be to stop the progress bar from appearing front and center on
> the screen, and put it elsewhere.


One other issue is that before applying the atomic system change, you  
need to verify the system is still in the same state as when the  
working copy snapshot was taken. Monticello works by comparing a  
snapshot of the package in the image with the package to be loaded,  
taking the difference to patch the system. But if the part covered by  
the snapshot is modified, the snapshot is invalid, and so the patch  
becomes invalid too.

- Bert -