xml-general mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Edwin Goei" <Edwin.G...@eng.sun.com>
Subject Re: [spinnaker] Announce
Date Tue, 11 Jul 2000 19:52:03 GMT
> The existing xerces-C scanner is a fundamentally different code base from
> xerces-j, at least in part because some of those tweaky Java optimizations
> seemed to dominate the design of xerces-j.

Yes, I think this is one area that makes the code difficult to understand
and is usually not necessary with newer VMs which is what I think the new
parser should be targeted for.  Current users can still use the current
version of Xerces.

> At some point XML schema will need to be done for C++, and I certainly
> hope that we [whoever actually ends up doing the work] will be able to use
> the design for schema support from xerces-J pretty much intact.
>
> Knowing that there will be both a Java and a C++ implementation of a given
> design does impose some constraints, but they're not too bad - a little
> extra thought on how memory management will work, don't rely too heavily
> on introspection and the like.  And it doesn't mean that the C++ code
> needs to come out looking like the Xerces C++ DOM, which was set up as a
> minimum effort port from Java.
>
> But if we don't think about the issues up front, when doing the initial
> architecture, we run the risk of having to do a complete redesign for C++,
> which would be a big waste of time and effort.

Yes, sharing the overall architecture might work OK.  I think it will depend
on the details.  I think if the code is written to be clear rather than
optimized for a particular JVM it would be easier to port to C++.

-Edwin


Mime
View raw message