cocoon-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Sylvain Wallez <sylvain.wal...@anyware-tech.com>
Subject Re: Status of XMLCompiler
Date Tue, 09 Jan 2001 15:36:15 GMT
Paul Russell a écrit :
> 
> * Davanum Srinivas (dims@yahoo.com) wrote :
> > Questions:
> > 1. Shouldn't we be Caching the output of the Serializer and not the
> > intermediate SAX events themselves? (For example the SVG Serializer
> > takes more time to process the request than the Generator's or
> > Transformers in the Chain)
> 
> There was some discussion of this some time ago, and the new result (if
> I remember rightly) was that we should cache both the serialized stream
> and *some* (but not all) of the intermediate sax pathways.
> 

At that time, I suggested the use of a new <cachepoint/> tag in the
sitemap to manually insert cache points in the pipelines.

IMO, automatic and sitemap-transparent caching may be difficult to tune
between memory space (useless caching) and computation time (costly
regeneration of non-cached streams).

<cachepoint> could allow a site admin to manually tune the cache in the
website based either on his knowledge of the site architecture or the
output of a sitemap profiler.

> > 2. Should we place the burden of deciding (whether the result of a
> > request should come from the Cache or generated again from scratch) on
> > the Generator? (This was the case with C1 right?)
> 
> I think it was decided that it was a little more complicated than this,
> due to pipelines joining through aggregation etc.
> 

Isn't <cachepoint> a solution to this problem ? I think we *must* have
caching in C2 before releasing it, and the design of a transparent cache
architecture can be lengthy. Manual caching can also help us to learn
how to best implement automatic caching in future releases.

My 0.02 euro.
-- 
Sylvain Wallez
Anyware Technologies

Mime
View raw message