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 17:58:18 GMT
> > > I'm a little surprised at the friendly infighting. Just for clarification,
> > > is Catalina destined to be Tomcat 4.0, or is there a chance that Tomcat 3.x
> > > could take that honor? I assume that this would come up to a vote amongst
> > > the contributors, but for those of us that are writing future tools and
> > > integration, it would be nice to know. When will this decision take place?
> > >
> > > I think it will come down to a case where there is no clear winner. It's the
> > > classic design vs. performance question. I would be surprised if the more
> > > componetized and better designed Catalina could outperform the more
> > > evolutionary, but harder to learn, Tomcat. Then again, this is supposed to
> > > be a reference platform, and design should win out.
> 
> To start,
> 
> 	I hate to sound like a hippie or something, but why can't both
> designs be good?  It would be extraordinary if someone was to determine
> that some design was inherently "right" in a case like this.  They are
> both too complicated to prove optimal by any mathematical means.

It's true, and that's why we need code to verify the claims.

In some cases, the differences are clear and can't be merged:
For example, Catalina choose to "inovate" by not using the Facade
pattern. Without facade you are constrained to the servlet interfaces,
which are may be best as a standard API, but it's not the best core API.

The only desing pattern that can be argued about is the Chain/Valve versus
Interceptor. As I said, it's easy to add Chain support in tomcat, but I'm
haven't heared any argument for that except it's "simpler and better
documented". 

I wouldn't mind having it - it's just that Craig wants a new code base,
and not tomcat, and he doesn't like the facade, adapters and few other
patterns used extensively in tomcat. Even if we add the valves to tomcat,
it will not change too much the situation.

> > But in the end the better design does matter - and that's where tocmat3.x
> > is (IMHO) better. Of course, you shouldn't believe me - use your own
> > experience, try to understand the patterns used.
> 
> 	At the minimum, I think Catalina is a great because its 
> different.  Sometimes the only way to make progress is to break out
> of a rut and see what works and what doesn't.

Agree with that. There are many ideas in Catalina. Of course, we can argue
that it's very bad that the implementation depends on a particular design
decision and it's not independent and reusable.


> 	It could be.  After all, it's possible that IIS and Apache 
> did it that way because NES had already done it.  I'm all for trying
> new patterns to see how they fly, as long as we are objective about
> moving back if necessary.

Right now it seems many want to move forward just because it's different.

Costin


Mime
View raw message