commons-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From robert burrell donkin <>
Subject Re: [digester][patch] xml schema support. removal
Date Sun, 19 Oct 2003 13:44:56 GMT
On Sunday, October 5, 2003, at 11:04 PM, Simon Kitching wrote:

> On Sun, 2003-10-05 at 07:26, Jean-Francois Arcand wrote:
>> Hi Robert,
>> robert burrell donkin wrote:
>>> hi Jean-Francois
>>> i'm very reluctant to apply a patch that breaks backwards
>>> compatibility but i'd be happier to deprecate them.
> That was my initial response too. However it seems that the current code
> is actually broken/unreliable. Is it worse to remove code without the
> usual deprecation stage, or leave known-broken code in a release?

we always deprecate. one of the reasons why jakarta-commons has a good 
reputation is the rule no code is removed with first being deprecated. 
this sometimes means an extra release but this is a small price to pay. 
(see later for the some reasons why users appreciate this process.)

> I'm not sure I would prefer to have an application fail to work
> correctly than to fail to compile...

when you're talking about backwards compatibility, the needs of the many 
often out weight the needs of the few. digester is now mature and is used 
(both directly and indirectly) in many applications. breaking backwards 
compatibility means that we will force users to recompile all these 
applications if they want any new bug fixes in future release. this 
process can often be very painful - or even impossible (think about 
proprietary software using a particular library version).

i did consider the alternatives. the existing code works for some versions 
of xerces - but not all. the users which are using this code must be using 
compatible versions. i've proposed an approach that should be able to 
retain backwards compatibility whilst also allowing a wider range of 
parsers to be supported. coupled with deprecation and documentation i 
believe that this will satisfy the needs of both new and existing users.

- robert

To unsubscribe, e-mail:
For additional commands, e-mail:

View raw message