cocoon-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Hannes Haug <Han...@Haug.com>
Subject Re: don't cache - compile
Date Thu, 27 Jan 2000 10:58:47 GMT
Niclas Hedhman wrote:
> 
> I was just reflecting on your statement "output only depends on input
> documents", which is incorrect.

It's correct for fop and for xhtml-to-html-printers. But
you're right. Some formatters might depend on e.g. accept
headers.
 
> The question with caching is how to figure out if a particular object has
> changed or not. And in a highly flexible architecture, I see problems with
> that if not every single stage involved in the chain is determining if a
> valid result object is available in the cache.

That works in my model, too.

> Maybe you are right in your assumption that;
> a) If you have no limits on how much documents are cached, and
> b) changes occurr near the beginning of the processing chain,
> you could achieve better performance with stage-validation of a final
> document, than try to be looking at small objects. I can however see a
> problem of how you know which is the final document prior to executing all
> the processors. But maybe I am thinking too far ahead.

In general you don't know the final document prior to executing
all stages, of course. But sometimes you know - in this cases
validation is enough. This has nothing to do with my proposal.

But my idea or question is: wouldn't a system based on inter
processor optimisation be faster than a system based on intermediate
document hashing? Since there is no real xslt compiler that can do
such things it's a rather theoretical question.

Once more: I don't know if my proposal would lead to an
efficient implementation. My thoughts are usually a request
for comment and not a request for implementation.

 -hh

Mime
View raw message