At 10:14 PM 3/4/2003, J Cameron Cooper wrote:
I think showing Zope to be faster than Tomcat, as the numbers indicate, will be difficult to support, especially since I don't believe it myself.
Although benchmarks are a very small part of the story as far as I'm concerned, there is a need for such numbers.
Indeed. Zope is a decent enough performer in benchmark testing and we should know where it stands. But the real case for Zope is not its speed as a server, but its speed as a development platform. You can increase performance speed easily enough... get faster hardware, a bigger pipe, more caching, a reverse proxy, do load balancing. It doesn't take a heck of a lot to build a world-class service *around* Zope. Where Zope shines is in time to market and ease of maintenance. It's going to be tough to give a precise benchmark on how much faster it is to build something in Zope than it is in J2EE or Jboss. But it is appreciably faster, perhaps by an order of magnitude. Developer time and support resources just aren't things most shops can come up with as easily as cash for hardware. The learning curve is shorter too. If you're a moderately experienced developer, you can pick up 90% of Python in a day or two. I've never heard that said about Java... at least not with a straight face. :-) Perhaps the best way to make a case for Zope is to prototype something. Give yourself a little time to do some reading up front, but then just rip into something. In many ways Zope is a gotta-see-it-to-believe-it proposition. Lots of platforms claim to make things easy... but the proof is in doing something nontrivial. In the time it takes to evaluate Zope and measure it against other platforms you could almost be done with a rough prototype of your system. HTH, Dylan