[jdom-interest] Toward beta 9

[Vadim.Strizhevsky at morganstanley.com]

Personally I would preffer your solution with optional Verifier. I also
don't believe that libraries should protect people from themselves at the
expense of efficiency. But I could live with compromise also.

I'm sure this is too late for beta 9(?), but can the resolution of this
issue one way or the other be placed for beta9+1 list. Since the original
discussion  was over a year ago (I guess I missed it), I feel its
important to put a solution for this. My usecases are really hurting by
the current overhead of the Verifier.

Thanks a lot,
-Vadim

On Wed, 9 Apr 2003, Alex Rosen wrote:

> We discussed this quite a bit a year ago:
> http://www.servlets.com/archive/servlet/BrowseList?listName=jdom-interest&by=subject&from=74264
>
> My philosophy was that protecting people from themselves is great when
> it's cheap, but in this case it's not. Elliotte feels that it should be
> impossible to create a document that's not 100% well-formed, even if
> that's expensive to check. (And to achieve that goal will get quite a
> bit more expensive - e.g. we'd need to read each referenced DTD and
> parse it, to make sure that entity referenences have been declared and
> are not circular.)
>
> My proposed solution was to make the verifier a separate, optional
> step. When the user creates a document in-memory (as opposed to parsing
> a file), then should run the verifier on it when they're done. Many
> people won't do this, and that's their prerogative, but at some point
> down the line some XML parser will find the error when their document
> gets parsed, so it's not the end of the world. (That's why the parser is
> required to be so strict, IMO - so the generator doesn't have to be.)
>
> Jason proposed a compromise solution: verify names but not values,
> which is likely to be relatively cheap. E.g. verify element and
> attribute names but not attribute values or text content. This is what
> DOM does. (dom4j doesn't verify anything, at least as of a year ago.)
>
> Alex Rosen
> Novell, Inc.
>
> >>> <Vadim.Strizhevsky at morganstanley.com> 4/9/2003 10:25:03 AM >>>
>
> Hi,
>
> I raised this issue recently about Verifier, but I don't think anyone
> responded. Currently Virifier runs both when you building the message
> "by
> hand" and when it comes from SAX parser. Howerver when it comes from
> SAX
> parser it is really unecessary, as it is XML parser responsibility to
> verify character set. All valid XML parsers including Xerces and
> Crimson
> do this. So Verifier is just doing the same work again and in a less
> efficient way and adds significant overhead. Is it possible that JDOM
> can
> do something so that Verifier is not used when message is constructed
> from
> XML parser?
>
> Thanks,
> -Vadim
>
> On Wed, 9 Apr 2003, Bradley S. Huffman wrote:
>
> > Elliotte Rusty Harold writes:
> >
> > > >Verifier check for ProcessingInstruction's setData()? Elliotte can
> we borrow
> > > >the check out of XOM's Verifier?
> > > >
> > >
> > > Sure. I officially grant the JDOM permission to republish that
> check
> > > under the JDOM license rather than the LGPL.
> >
> > Got love open source :)
> >
> > Hmmm, it looks like this is the last set method that didn't have a
> call
> > to a check in Verifier. After this check is added, does that cover
> all
> > the XML spec. constaints that can be checked in Verifier and can we
> cross
> > that item off the TODO list?
> >
> > Brad
> > _______________________________________________
> > To control your jdom-interest membership:
> >
> http://lists.denveronline.net/mailman/options/jdom-interest/youraddr@yourhost.com
>
> >
> >
>
> _______________________________________________
> To control your jdom-interest membership:
> http://lists.denveronline.net/mailman/options/jdom-interest/youraddr@yourhost.com
>
>