cocoon-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Zvi Avraham <>
Subject Re: Variations on a theme by Cocoon
Date Tue, 15 Feb 2000 08:23:36 GMT

Niclas Hedhman wrote:

> Zvi Avraham wrote:
> >
> > Sounds OK to me, but I wandering if possible to "linearize" entire path of
> > producer -> filter -> serializer
> > or
> > producer -> filter1 -> filter2 -> ... -> filterN -> serializer
> > compile all this into one .class file (Servlet) ?
> >
> > Then you even don't need Cocoon in runtime, only in design time (compile
> > time? :)
> This is not a clear cut case.
> First of all, this is already possible. Generate the Site content into static
> html files and have them served by the web server. Won't get faster than that
> in runtime.

That case can be easily recognized too.
If in producer stage, we have static XML file (with no or static entities
included), then is possible just to apply all transformations (again if all
conditions and parameters are static too) and write the result of all
transformation as output .html (or .vrml etc..) static file.

> But, that is not the issue.
> If you have the pipeline above, one must evaluate the "cost of validation" CoV
> and "cost of regeneration" CoR versus "frequencies of change" FoC (dynamic
> content for instance) versus "Cost of Transport" CoT.
> If you are using smaller stages, the FoC will drop, CoV will be roughly the
> same and CoR will be reduced. CoT will be increased due to more transporting
> stages, but possibly a lot lower with SAX.
> One should possibly also introduce "Regeneration Latency", which will increase,
> possibly exceeding "Maximum Allowed Latency".

Too much math :)

> This is a pretty complex formula, but I think it would be safe to say there are
> no clear and simple answers. It all depends on the scenario of static versus
> dynamic and semi-dynamic content, request frequency, content volume and other
> factors.
> Niclas


View raw message