[Expat-discuss] Expat-discuss Digest, Vol 72, Issue 7

Karl Waclawek karl at waclawek.net
Fri Mar 10 04:33:03 CET 2006


Jimmy Zhang wrote:
> Buffer reuse is introduced in the latest releases, maybe it should always be 
> used, the
> performance improves starts from the second time VTDGen parses XML 
> documents..
>   
The same document or any?
> We are adding more documents on this feature right now...
>
> External references not withstanding, VTD-XML conforms to, and passes, every 
> XML
> test suite, 

I recommend to put a script or small program on your web-site that runs 
vtd-xml through the official xml test suite.
This will make potential users gain confidence with it. The SAX for .NET 
project has such an application,
you can use it as a blue-print. The mere fact that after 10 minutes of 
trying I already discovered a problem does
not bode well for your claim that VTD-XML conforms to, and passes, 
*every* XML test suite.
> VTD-XML handles namespace problem a little different than DOM or 
> SAX,
> the error checking is delayed until during navigation, the prefix induced 
> attribute duplicate
> problem is quite unlikely to concern anyone, and is in fact part of the 
> problems of XML
> namespace spec...
>   
Well, being 99% correct is not good enough.
> The cost of encoding transformation ranges from zero to negligible, most are 
> ascii anyway.
>   
Depends on the amount of character data the application retrieves.
> One can argue that, to process XML, SAX parsers need to be used at least 
> twice...
> first time is to scan the document from start to end, just to check 
> wellformedness, the
> second pass is to perform the application processing... otherwise, what 
> happens
> if the application perform 10 transactions but then discover that  the last 
> angle
> bracket of the XMl file is missing.?? roll back those 10 transactions ??
Yes, that is what transaction processing allows you to do. Or do you 
suggest not to use transactions
just because you know the XML is well-formed?
>  So 
> should we
> reduce the SAX perform by 50% just to be fair comparison with VTD-XML??
>   
That argument makes no sense to me.
> and VTD-XML is still forward only and unpleasant to use...
>   
But especially good at processing large giga-byte sized data dumps. 
Expat is routinely used  for such purposes.
> I don't see any comparison...
>
> Maybe the world has moved forward... maybe it is time to say good bye to 
> SAX??
>
>   
For some applications, I agree.

Karl


More information about the Expat-discuss mailing list