cocoon-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Joerg Heinicke <joerg.heini...@gmx.de>
Subject Re: Can I apply COCOON-2145.patch
Date Mon, 26 Nov 2007 05:03:16 GMT
On 25.11.2007 10:32 Uhr, Grzegorz Kossakowski wrote:

>>> I added an issue with a new input module which returns the full file
>>> path location of the base dir/resource of a block.
>>>
>>> https://issues.apache.org/jira/browse/COCOON-2145
>>>
>>> Can I apply the patch?
>> Sorry but I've no idea - I'm not clear yet on blockcontext thing myself!
>> :) Reinhardt? Grzegorz? Can new expression stuff be of help here?
>>
>> I'd like to see better name though... 'blockpath', may be? Or
>> 'blockcontext'.

> I'm quite reluctant to accept this patch because I believe that one should never rely
on block's
> directory location

Why? It might be necessary when communicating with other systems. There 
is no difference than with WAR and servlet context directory.

> However, I don't have time to really evaluate Thorsten's use-case now.

Isn't that what we have a list for instead of single people :-) If he 
has not done so far Thorsten can provide a short summary of his use 
case. Otherwise please point out the mail in the archives.

> I would say that I'm ok to add this input
> module but with fat warning that it should be used only if there is no other way and
we are *not*
> going to support compatibility of blockpath module in future versions.

I consider this bad idea. This creates just uncertainty for the user. 
Either add and support it - or not. We still could point to that issue 
and say: If you really need it get it from there.

My concerns go into a different direction. It might be necessary to know 
about the actual file system path as mentioned above. But it should not 
be necessary for the code running in a block to be aware that it is 
running in a block. The block is a way of encapsulating functionality, a 
web app is another one. This should be an abstraction and not known 
"inside". Why don't we reuse context:? This would make the code 
independent from the way its packaged.

Joerg

Mime
View raw message