On Sun, Jul 20, 2008 at 11:00 AM, <[EMAIL PROTECTED]> wrote:

> Author: ieb
> Date: Sun Jul 20 11:00:48 2008
> New Revision: 678290
>
> URL: http://svn.apache.org/viewvc?rev=678290&view=rev
> Log:
> Unbound the DataServiceServlet BeanConverters from their implementations
> and rebound them to the
> Interface. Modified the injected setter to use named beans and updated the
> relevant modules to inject the same
> implementations as before.
>
> This change will allow alternative implementations of a BeanConverter to be
> injected by Guice, and avoid binding
> to implementations. There is no functional change created by this patch.
> About 10 lines in total.
>
> I am assuming commits like this, which dont generate functional change  are
> small enough not to require
> JIRA discussion, but please tell me if this is not the case ?
>

Our rule of thumb has been that a trackable discussion (JIRA or shindig-dev,
with JIRA strongly preferred) is only necessary if any of the following
apply:

- It changes an intentionally public API (i.e. most interfaces) in an
incompatible way. Note that we have many unintentionally public APIs, but
fixing those will take more time.
- It's a substantial amount of code (say, more than a few dozen executed
lines)
- It adds a new dependency.
- It involves anything that requires a substantial amount of explanation to
understand (hacks, platform specific code, and anything with quesitonable
security implications falls into this category).

When in doubt, create the ticket to give people a head's up. Use your own
judgement for determining the impact of the change, which in turn should
determine how much discussion it needs.

Mostly we want to avoid big surprises, especially as we're getting closer to
release time.

Ensuring that thorough tests for your changes are included will also make
the change more palatable. If someone makes a change that breaks your code
and you didn't have tests, it's your fault, not theirs.

Reply via email to