----- Original Message -----
From: "Dmitri Colebatch" <[EMAIL PROTECTED]>
> > > Why cache to XML if you're looking for super performance?  Wouldn't
> > > serialized objects be even quicker?  I haven't a clue how xjavadoc
works
> > > internally, but my hunch is that it builds an object graph
representing
> > > classes/methods/fields and their @tag data.
> > >
> > > Just an idea.... dunno if loading serialized objects is faster than
SAX on
> > > XML files and rebuilding the object graphi, but seems like it would
be.
> > >
> > >     Erik
> >
> > Hmmm. Serialized objects were used by NetBeans in the early days to
persist
> > the state of the IDE (user settings, window positioning etc.) and this
was
> > not exactly successful since new versions of classes are persisted
> > differently and incompatibly...
>
> but thats configuration information yes?  In this case, there would be no
problem if the cache was corrupt... we can catch the
> exception (error?  geez... hmmm) and dump the corrupt part of the cache
and build again...
>
> cheers
> dim

True, of course! The serialization was used by NetBeans to store lots of
configuration settings and even project settings, things that are done
mostly manually. Every new release was a PITA to upgrade to!

XJavaDoc generates all info that is serialized, so there are no issues here
WRT compatibility If compatibility issues arise XJavaDoc would simply delete
the 'old' serialized database and run a full parse.
However, it might be interesting to use XML as persistency format just to
open up XJavadoc to other, non-xdoclet tools (like generating docs using
XSLT (actually do anything with XSLT!) or some validation).

The most important driver to which persistency format should be used IMHO
would be performance.

Now Thinking Out Loud:
The other driver, not much less important, would be usability! Like I said,
using an XML format certainly opens doors to many other tools which process
XML like XSLT and Schema validation. Hence, XML should be the number one
exchange format between XJavaDoc and XDoclet! Every XDoclet would simply be
a SAX ContentHandler and can do the same things they're doing now
(generating descriptors, html, sourcefiles) or/and other SAX ContentHandlers
(XSLT, validation).

Is this just plain stupid or might this actually work!?

ciao
BG



_______________________________________________
Xdoclet-devel mailing list
[EMAIL PROTECTED]
https://lists.sourceforge.net/lists/listinfo/xdoclet-devel

Reply via email to