Re: [xslt] xsltproc memory consumption w/ large DTD / docbook

[Daniel Veillard <veillard redhat com>]

On Sat, Feb 01, 2003 at 05:41:52PM +0000, Sebastian Rahtz wrote:
> >   DocBook stylesheets also use unparsed-entity-uri() and that
> > requires knowledge of the attribute type defined only in the DTD.
> 
> surely not so?

  I stand firm on this. 

> a simply well-formed document <x foo="bar"/>
> can have an XSLT process it under the assumption that "bar" is
> an entity (defined in the DTD subset)

  and need to map back to the system ID for that entity hence requiring
the entity declaration from the DTD. Please recheck the spec, I'm 95%
sure about this (but there is those 5% ...)

> a third technique is to have a single virtual document which
> pulls in all the others using XInclude. that way the DTD only
> gets read once.

  No, it will be read for every instance, even if XIncluded. As long
as the separate chunk references the DocBook DTd (and they do !) 
XInclude requires to at least load the DTd for entity location.
The difference is that in the XInclude case the document will be discarded
after the inclusion got done so the N copies of the DTD will not
be kept in memory at the same time.

Daniel

-- 
Daniel Veillard      | Red Hat Network https://rhn.redhat.com/
veillard@redhat.com  | libxml GNOME XML XSLT toolkit  http://xmlsoft.org/
http://veillard.com/ | Rpmfind RPM search engine http://rpmfind.net/