Thoughts on moving objects, XML and namespaces etc...

Marcel Weiher marcel at system.de
Tue Jul 27 11:51:35 UTC 1999


> From: "Gàran" Hultgren <gohu at rocketmail.com>
>
> 4. XML is very nice. I work with it in Java and we are using it right 
> now both as a more robust way of representing objects in files than 
> serialization. If you do it right and couple XML with some nice
> "builder" objects you can isolate your testdata better from the
> evolving class schemas.

You can also combine serialization with specific encoders to get the  
best of both worlds.

> Another perfect fit for XML is representing UIs as Mozilla are doing 
> in their Xul stuff. VWs windowsSpecs together with builder objects 
> were in retrospect "the right way to go". :-) I hate UI code
> generation and I do not really believe in UI serialization as Sun
> seems to do. And hacking the UIs by hand is tedious and results in 
> code bloat. Go XML.

Yes, NeXT's Interface Builder always used 'freeze dried' objects and  
this was one of the few things that were (and still are) actually  
better than the Smalltalks of the day.  Apple is also going to change  
the format to an XML encoding.

> But serialization of objects between languages is hard. HARD on the 
> brink of impossibility I would presume. OMG/CORBA are trying aren't 
> they? I say good luck. :-)

In a way, CORBA/IIOP does exactly that.  It 'serializes' calls,  
which are really just instances of class Message, after all. :-)  The  
mechanisms they have (encodings for basic + composite values +  
proxies) pretty much cover what is needed.  One of the problems is  
that the encoding isn't typically self-describing, so you have to  
agree on exact formats beforehand and cannot deal with any  
deviations.  Another (IMHO) is that IIOP is yet another specific  
'serialization' byte-format.

Marcel





More information about the Squeak-dev mailing list