I thought I'd just post here what I sent to the axdtk list, in case there
are people here who don't know what the heck I'm talking about... If this
perks your interest, send a mail to [EMAIL PROTECTED], download
the toolkit, and start asking questions...
What is the AXDTK?
It's a system very much like Cocoon (http://xml.apache.org/cocoon/), but
based on mod_perl. The idea is to allow a few things:
- Stylesheets and XML files get associated automatically using the w3c
semantics (http://www.w3.org/TR/xml-stylesheet).
- Pre and Post processing can be performed using mod_perl stacked
handlers.
- Output is cached automatically if the underlying system uses DOM,
otherwise the stylesheet processor developer has to implement caching
himself, however the hooks are provided for this.
- Stylesheets can cascade if based on DOM trees.
For those unfamiliar with Cocoon, here's what happens:
1. A request comes in for an XML file.
2. If someone has installed a stacked handler prior to
Apache::XMLStylesheet, that handler is called. One idea for such handlers
is to extract a media type by browser sniffing (e.g. sets the media to
"handheld"). Another idea is to take a preferred style from the
querystring such as "style=no%20images".
3. Apache::XMLStylesheet checks its cache for a matching set of combined
parameters (file, style, media). If the cache says that the XML file is
older than the cache, use the list of stylesheets in the cache. If the
cache is older than the file, parse the file for the xml-stylesheet
processing instructions.
4. Check all the stylesheets exist.
5. If all the stylesheets exist, check the difference in timestamps
between the xml file and all the stylesheet files, and the output cache
file. If any are newer than the output cache file, then we re-run all the
styles against the xml file. If they're all older we stop there, set the
filename to the cached file, and return DECLINED. Apache obliges and
returns the cached file to the browser for us.
6. If we re-create, we have mark off stylesheet type mappings against
modules or functions. That is, the type="..." attribute in the stylesheet
processing instruction gives us a pointer as to what module to use to
process that stylesheet (based on some info in your httpd.conf file). We
load that module and execute the function, passing in the stylesheet and
the xml filename as parameters.
7. Modules based on the DOM can use the mod_perl API to set the
$r->pnotes('dom_tree') value when they're done, rather than printing out
the results. This is how the caching works, and how cascading
works. Modules that aren't based on the DOM can't cascade... yet.
8. The resulting DOM tree is printed out both to the cache file and to the
browser. The DOM tree is disposed of.
That, in a nutshell, is it. Currently it's a work in progress. Today I
hope to finish the version that does all the caching work and cascading
stylesheets.
If anyone wants to volunteer to help, here's what I consider needs done:
- A cool web page ;-)
- Get Xalan/Xerces working on Unix. This is pretty vital to the success,
IMHO. See http://xml.apache.org/
- Write modules that plugin to this architecture that perform the
following functionality:
- SQL plugin using DBI. Should take an XML file (or DOM tree), parse
out connection information and run the SQL. Produce another DOM tree and
put that DOM tree in $r->pnotes('dom_tree')
- Implement other stylesheet processors, like XPathScript and NotXSLT
- Get the word out! Perl must not be behind in the XML world for much
longer!
Download AXDTK at http://xml.sergeant.org/download/
It's not well organised right now. The key component is
Apache::XMLStylesheet. To use that you need XML::Parser and
Apache::MimeXML(0.05). The only plugins available are Apache::XPathScript
and Apache::XPath::NotXSLT, and an XSLT plugin that's in the list
archive.
--
Fastnet Software Ltd. High Performance Web Specialists
Providing mod_perl, XML, Sybase and Oracle solutions
Email for training and consultancy availability.
http://sergeant.org http://xml.sergeant.org
