geronimo-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Aaron Mulder <ammul...@alumni.princeton.edu>
Subject RE: [PATCH] [JSR-88] DFM Patch Take 3
Date Sat, 16 Aug 2003 16:45:06 GMT
On Sat, 16 Aug 2003, Maas van den Berg wrote:
> However the RI passed the certification tests (I assume) therefor
> mimicking its behavior won't hurt us.

	Define "hurt us".  If you only goal is to pass the TCK, then I
agree.  But my goal includes having a good implementation as well.  
Searching a List with several nulls and several copies of the same object
does not strike me as a particularly good implementation.  Preventing the
possibility of a user ever seeing the message or stack trace of an
exception that was generated also seems pretty dubious.

> Also javadoc/specs just don't capture all requirement, you'll always
> need to test against the RI to discover the details.

	Are you saying that in order to code against the API, you can't
use the JavaDoc or the spec, but you must write code that takes advantage
of specific features in the RI?  I disagree.  I've never used the EJB RI,
but I've written loads of EJB applications.  If I were to take advantage
of an unspecified, undocumented implementation detail in the RI, and that
caused my code to fail in WebLogic or Geronimo, whose fault would it be?  
Theirs for not exactly duplicating the implementation of the RI?  Mine, I
think.

Aaron


Mime
View raw message