tomcat-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From <cos...@eng.sun.com>
Subject Re: Tomcat, Catalina, and Java2
Date Tue, 11 Jul 2000 17:04:51 GMT
> I'd like to voice some concern as well.
> It's confusing to me.
> Should I be architecting my session server for Catalina instead of Tomcat?
> Is the time I've invested in learning how Tomcat works wasted?

Learning is not a waste. Read and understand Catalina, Jigsaw, Apache,
IIS, NES - it's all usefull. 

Then decide for yourself what model is better, what patterns you need. 


> Sure, Tomcat has it's flaws, but they don't seem terminal.
> 
> I really like that Costin appears to be integrating the "good stuff" from
> Catalina.  If the best features of Catalina are assimilated, why have a
> Catalina?  Why not work to improve the Tomcat collective?

It is possible to integrate Valves, but that will provide little benefit -
Interceptors are much better from many points of view, and that's even
more important because our goal is to work with web servers, which use a
model similar with Interceptors. If enough good valves will be implemented
it is of course possible to include this pattern in tomcat, but right now
it's not worth the effort.

About the core catalina interfaces ( which are the write 1/2 of servlet
interfaces ) - that's completely impossible, it prevents two very
important goal of tomcat - security and performance. It's very hard 
to achieve security if you don't have a clear demarcation, and it's almost
impossible to achieve good performance if you are constraint by a
"user-level" interface in your core.

I'm not talking about tomcat 3.2 performance - fact is that all you see in
3.2 is just the result of cleaning up the garbage. The real performance
will come after 3.2, when we can start to take advantage of the facade
architecture.

What else? I think the TCP code is far better in tomcat ( and it's easy to
verify that - we support 5 protocols, with more to come ), and I would be
very -1 against changing it without some serious reasons.

The session will be integrated sooner or later, same for the http adapter.


> 
> History shows the best architecture/software design most definitely does not
> win by default.

I certainly hope it will. At least in open source world. 

Costin


Mime
View raw message