[jsr-314-open] My performance AI from the EG meeting

[Jim Driscoll]

So, I've been thinking about measuring relative performance, and I think 
that what we may want to do is to try to get JSF into the next specweb 
standard.  There are currently ASPX, PHP and JSP in the standard. 
http://www.spec.org/web2009/

We'll see what Sun's performance guys say.

That would allow us to 1) compare JSF implementations for performance 
(always useful), and 2) compare JSF releases for performance, to guard 
against performance degradation from spec changes (very, very useful).

We'll also want to port over some subset of the tests to something like 
Wicket.  It also might be fun/interesting to port over some tests to 
Rails as well.  I expect that Wicket may have some advantage in some 
cases, since the programmer creates the tree, rather than the framework 
- but then, I can also write things faster using an assembler, for much 
the same reason, with some of the same problems.  I also expect that 
we'll find a few very obvious places to boost speed - we've been mostly 
striving for correctness in 2.0.

Ajax tests, otoh, are likely to be much, much harder.  There are 
commercial products, like the neotys.com one that Ted mentioned, but I 
don't have to tell you that my budget for this is $0 - maybe that can 
change, but I suspect I'll be writing client scripts and using Selenium. 
  Ick.

Anyhow, it's a long term project, and I wanted to update you on my 
thinking, and seek feedback.

Jim