On Fri, 11 Jan 2002, GOMEZ Henri wrote:

> Unix domain will be faster on systems running Apache and Tomcat on
> the same box, so faster network IO will still be fine for configuration
> with separate Webserver and Tomcat (security concerns or load-balancing)

Of course, different issues. For 'lb' you have an extra layer of
scalability - you can add more workers to better handle  the load.
Assuming a normal case, with servlets doing something besides helloworld,
the benefits for NBIO on lb will remain small.

As I said, I'm +1 on NBIO ( and I love the new java.nio package ), not
because of possible improvements on tomcat's performance  but because it's
cool and the right ( or better ) way to do io.


> >Regarding NBIO and Jk2 - I think the Channel abstraction could be
> >enhanced to use poll() ( using APR ).
>
> Jk2 is really a whole new study for the good old mod_jk developpers,

What I tried to do is make it easier to read and cleaner - for
new and future 'old mod_jk developers'.

There is a big change in code organization - but if anyone feels the
code is harder to read and maintain then we need to start over
again. I personally believe it's going in the right direction,
so I'll continue moving.

BTW, the new configuration ( based on Location instead of JkMount
and without workers.properties ) needs review and feedback.
( the old stuff should still work, but I would like it deprecated )

Costin


--
To unsubscribe, e-mail:   <mailto:[EMAIL PROTECTED]>
For additional commands, e-mail: <mailto:[EMAIL PROTECTED]>

Reply via email to