Re: A question @Mark Wilaard (and other developer)

[Audrius Meskauskas <audriusa@xxxxxxxxxx>]

Chris Burdess wrote:
> Yes. This vexes me greatly, when there is pressure from the community 
> to reproduce a bug-for-bug compatible class library, i.e. that we 
> should try to code to match Sun's implementation instead of coding a 
> good, bug-free implementation of the specification. But, as you point 
> out, in many cases the specification is vague or nonexistent, and 
> people just want it to work and don't care about specs.
For sure. I have the detailed tests for both our HTML parser and our 
CORBA implementations. Only part of these tests is committed to Mauve, 
as the remainder fails tragically with Sun's distribution and is not 
very desired there. These tests, however, are checking things that MUST 
work following the official OMG and W3C specifications, and in the case 
of CORBA they are not invented and written by me alone. I still silently 
check for regressions against them, despite if somebody else would start 
actively breaking such functionality, it would be difficult to argue.

Under circumstances when near any Sun's bug is a desired behavior, the 
question "in which part Classpath implementation is better" may look 
just inappropriate.  I really hope that after open sourcing Sun's java 
under acceptable license and with several implementations in competition 
such bugs may finally loose the undeserved status of the "unofficial 
standard".

Audrius