geronimo-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Adrian Jackson <Adr...@extenuating.com>
Subject Re: Status: JavaMail API implementation now at Alpha quality
Date Wed, 17 Sep 2003 10:43:59 GMT
Bill de hÓra wrote:
> [[[
> 1) Implement what the spec says
> 2) If the spec is ambiguous, talk to the spec lead
> 3) If you think the spec/lead is is wrong, do what the RI does
> 4) If the RI has a bug, file the bug, but stay bug compatible
> ]]]

Trying to stay bug compatible with a reference implementation is a 
nightmare: it's like trying to write code that conforms to a spec you're 
not allowed to see that changes from version to version without 
notification.

I guess the real question that our theoretical friends haven't been 
addressing is whether the implementation of the equals() method is 
actually something of any real practical importance. Is there software 
out there that's likely to run into the situation where the question of 
whether the equals() method for URLName is transitive in certain cases 
where different subclasses of URLName are compared becomes an issue? If 
not, then I'd prefer to see a 'correct' implementation as proposed by 
Alex. If it *is* an issue, and software is likely to depend on the 
'bugged' RI, then Jeremy's version of the code is a lot more appealing - 
practicality has to override theory at some point. Although some 
notoriously stupid bugs have been perpetuated by a refusal to fix them 
and potentially break code that relies on the bug in the first place 
(Java bug 4030718 springs to mind as an example). And as has already 
been suggested, ensure the bug in the RI has been reported to Sun.

But what do I know? I'm only a lurker round here. :)

Adrian



Mime
View raw message