RE: [logging] DON_QUIXOTE branch

[Andy McBride <an...@dsl.pipex.com>]

This is good to see, anything which leads towards a more predictable
behaviour for CL is great.  

I would personally prefer the static binding approach over byte code
manipulation as a more deterministic approach.  

I'm curious at to your choice of 'Don Quixote' as a branch name, please,
please put me out of my misery and explain what the reference means! :-)

Regards,

Andy 

> -----Original Message-----
> From: robert burrell donkin [mailto:robertburrelldonkin@blueyonder.co.uk]
> Sent: 31 March 2005 21:57
> To: Jakarta Commons Developers List
> Subject: [logging] DON_QUIXOTE branch
> 
> i've committed some code into a branch called DON_QUIXOTE. it's
> illustrative code showing how it's possible to lift off a simple
> superclass from LogFactory. i've believe for some time that this is the
> most important step forward. moving to a simplified API would allow
> static binding (whether compile time ala UGLI or byte code manipulation)
> to be offered in addition to (improved) dynamic binding.
> 
> one aspect that has been holding me back is the increased complexity
> that this choice gives. however, i'm now convinced that it really isn't
> any use running away from the complexity: we need to cover advanced use
> cases with better documentation as well as giving some intermediary
> heuristic recipes (to stop JCL blowing up so much).
> 
> this kind of design is (i think) one way forward for JCL. opinions
> welcomed (but please forgive the implementation: it's only
> illustrative).
> 
> - robert
> 
> 
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: commons-dev-unsubscribe@jakarta.apache.org
> For additional commands, e-mail: commons-dev-help@jakarta.apache.org




---------------------------------------------------------------------
To unsubscribe, e-mail: commons-dev-unsubscribe@jakarta.apache.org
For additional commands, e-mail: commons-dev-help@jakarta.apache.org

RE: [logging] DON_QUIXOTE branch

[robert burrell donkin <ro...@blueyonder.co.uk>]

On Fri, 2005-04-01 at 00:47 +0100, Andy McBride wrote:
> This is good to see, anything which leads towards a more predictable
> behaviour for CL is great.  
> 
> I would personally prefer the static binding approach over byte code
> manipulation as a more deterministic approach.  

both kinds of static binding (byte code and compiled) are equally
deterministic. both make it easy to predict which logging system will be
used. there are a number of advantages that byte code manipulation
offers including direct access to the logging system (for people like
simon who worry about performance) and the ability to dope particular
libraries to log to particular systems. compiled static is easier to
understand and explain, though.

note that dynamic binding is also deterministic: it's just that a lot
more knowledge is required to determine the behaviour in a given
situation.

the main advantage of dynamic binding is that can (potentially) function
in situations where static binding cannot. the same shared library can
also log to different log systems depending on context. for more
details, see the analysis code in SVN.

IMHO the best approach would be to allow a choice of approaches.

> I'm curious at to your choice of 'Don Quixote' as a branch name, please,
> please put me out of my misery and explain what the reference means! :-)

http://en.wikipedia.org/wiki/Don_Quixote  

- robert


---------------------------------------------------------------------
To unsubscribe, e-mail: commons-dev-unsubscribe@jakarta.apache.org
For additional commands, e-mail: commons-dev-help@jakarta.apache.org