> Barry A. Warsaw writes:
>  > concerns.  After talking with Fred, I came to understand that much of
>  > the code's current obscurity is due to optimizations to make TAL
>  > perform at least as fast as DTML.
> Well, faster than it used to.  We need to spend time on performance.

Indeed. All my "pseudo-benchmarking" resulted in ZPT being up to 4 times
slower than DTML ...

Wouldn't it possible to precompile ZPT into Python bytecode or even
something more efficient? We've compared DTML and ZPT to working with Python
strings that get variables passed, and the later was way faster. Which is
more or less obvious as it does not do nearly the same DTML/ZPT do (i.e.
security checks and the like) ...

I'd probably be happy with some way of creating pseudo-dynamic templates
that only have dynamic "hot spots", like server side includes in Apache, but
use precompiled or cached version of the template for the rest. The problem
is that most of the time we could use Zope + a cacheing proxy, except for
very small parts of the page that make it uncacheable as a whole ...

>  > So while I'd love to have the code be cleaner and more pluggable, we
>  > have to decide 1) if we're willing to give up some performance to
>  > acheive this, and 2) if not, can you provide a design and
>  > implementation that will "perform adequately"?
> It might be time to investigate Pyrex:
>     http://www.cosc.canterbury.ac.nz/~greg/python/Pyrex/

This one sounds good ...


Zope-Dev maillist  -  [EMAIL PROTECTED]
**  No cross posts or HTML encoding!  **
(Related lists - 
 http://lists.zope.org/mailman/listinfo/zope )

Reply via email to