[Portland] PortlandVFSProposal

[nf2]

Martin Konold wrote:

>Am Samstag, 7. Januar 2006 10:21 schrieb nf2:
>
>Hi,
>
>  
>
>>If the
>>layering would have been done correctly there wouldn't be a necessity
>>for project portland, RuDI,...
>>    
>>
>
>Portland and RuDI have a much larger scope than only file i/o. 
>  
>
Of course. Common infrastructure as well. What i meant is that the 
majority of the PortlandIntegrationTasks could also be solved with 
common infrastructure. And the infrastructure approach might be nicer, 
because it has a clear hierarchy of dependencies. KIO and Gnome-VFS 
would for instance just depend on the Common-VFS package.

With the "adapter approach" (RuDI) you always have to deal with odd 
cases like "there is no VFS library installed". Because of the 
late-binding. It's a wild mix of layers, the most abstruse example is 
probably this out-of process popping up filedialogs.

>In general I believe that standardizing on interfaces or better protocolls is 
>much better with regards to future innovation compared to standardizing on 
>implementation.
>  
>
Infrastructure libraries can also be improved/replaced behind a 
standardized interface. They could even serve multiple major versions of 
an interface (by providing compatibility stubs - using differend 
namespaces for function names).

>BTW: Is your proposal synchronous or asynchronous?
>  
>
both.

>Regards,
>-- martin
>
>  
>
I wouldn't say that the "adapter approach" is bad, but some things 
(out-of-process file- or printdialogs) look quite frightening...

Regards,
Norbert