> I don't know all the context - but I was pretty convinced that BSF is a
> good solution for script enabling JSPs and anything else if you want to
> support multiple languages.

To be clear, I'm in agreement with you.  Bill Stoddard was the one arguing
otherwise.  :-)

> > Benchmarking, as I've done recently, Apache -> mod_jk2 (UNIX domain
> > sockets) -> Tomcat 4.1, shows the somewhat expected *huge* difference in
> > performance between httpd and tomcat.

> When serving static requests. Remember tomcat is a servlet engine, it's
> job is to serve dynamic content, not static pages, that's Apache's role
:-)

I concur, which is why I have httpd doing its thing, and mod_jk2 taking
dynamic requests to tomcat.  You may have misunderstood my point, perhaps
because my premise wasn't accurate based on your data.

The point I was trying to make was that the overhead of handling a (simple)
JSP page request is significantly higher than the execution time of the
service() method, and therefore the additional BSF overhead might not be as
visible to users.  You seem to be saying that for Tomcat in a standalone
mode, the service() method is the largest part of request handling, even for
simple pages.

> 4 ms per request is pretty small, it would be smaller with tomcat
> standalone

I wasn't being critical.  Just empirical.

> For a simple JSP page - the execution time is the same with the execution
> time of a servlet, and comparable with that of a apache module. I did a
> lot of tests ( with tomcat standalone )

You figure that the overhead of mod_jk2 was 3-4ms per request?  Good to
know, both in terms of the overhead of the connection, and in terms of
internal performance with Tomcat.  I should benchmark Tomcat standalone at
some point this month, but I've no plans to run standalone in production.

As I said, I'm in favor of seeing BSF support incorporated directly into
Tomcat to implement the language attribute, and I don't see any alternative
that is more favorable (which isn't to say that BSF can't be further
optimized).  And, as you say, with a good connector dynamic content can be
distributed to multiple workers as load increases.

        --- Noel


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

Reply via email to