Re: svn commit: r665954 - in /cocoon/branches/BRANCH_2_1_X: ./ src/java/org/apache/cocoon/environment/ src/java/org/apache/cocoon/environment/wrapper/ src/java/org/apache/cocoon/util/ src/test/org/apache/cocoon/util/

[Joerg Heinicke <jo...@gmx.de>]

On 09.06.2008 22:40, joerg@apache.org wrote:

> Author: joerg
> Date: Mon Jun  9 19:40:16 2008
> New Revision: 665954
> 
> URL: http://svn.apache.org/viewvc?rev=665954&view=rev
> Log:
> as a follow-up of COCOON-2168 (http://marc.info/?t=120899868500003&r=1&w=4):
> Cocoon's pipeline buffer increases from an initial buffer size of 8192 bytes to the configurable flush buffer size rather than allocating the complete buffer beforehand.

I'm still working on making initial buffer size configurable in sitemap.

> +    public OutputStream getOutputStream(int bufferSize) throws IOException {
>          // This method could be called several times during request processing
>          // with differing values of bufferSize and should handle this situation
>          // correctly.

> +        // FIXME (JH): Question is what "correctly" means. The current behavior
> +        // seems to be inconsistent: On a second call with bufferSize == 0 we
> +        // discard whatever the first called set up. With a bufferSize != 0 the
> +        // first call's setup is preserved. Why not always creating new
> +        // BufferedOutputStream in the else block replacing a potentially
> +        // existing one?

Any idea/comment on this? If the current behavior is correct or expected 
it should probably commented appropriately.

Joerg