RE: [eclipse-dev] Please help me stop the context menu abuse

["Walter Harley" <eclipse@xxxxxxxxxxxxxx>]

Max Rydahl Andersen responded to my comments about UI changes:

> I have not yet heard *anyone* mention a concrete adopter or 
> user saying that this change would be considered bad ? 

As one concrete data point: BEA does have GUI-based automated test
scripts, that will probably be broken if context menu items get removed.
I no longer work at BEA so I have no way of determining what the
dependencies are; and perhaps, with Oracle acquiring BEA, it's moot.
But it's an example of the sort of breakage that is likely to occur.


> If they don't want to adopt they use the same "API" as 
> before, please tell me what is bad about this ?

But the problem is only solved if teams do adopt it, right?  You seemed
to be wondering why a team wouldn't adopt when the cost of changing a
menu item seems so low.  I was trying to explain that the cost is not in
the dev work itself, it is in figuring out how to assess and manage the
risk.


> I feel it is weak that the PMC does dare to do such a small 
> change that will make a big improvement.

I mentioned earlier that Platform may not be feeling the same pain, on
this issue, that downstream feature teams are.  Conversely, I think
downstream teams might not be aware of some of the pain that Platform
has been facing.  The last couple of months have been rather rocky for
the Platform, with the introduction of p2 and the API tooling and with
significant hardware and software problems in the build/test lab.  I
think the PMC is probably right to want to stabilize aggressively now.

As an example of something that seemed like a safe, innocuous change,
last week we had a problem where the new API tooling noticed that a
bundle had its minor version incorrectly advanced during 3.4, so the
responsible team corrected the version number back to what it should be.
The resulting downstream breakage wasted several days for WTP (and David
Williams can't afford to tear out very much more hair).

RC1 is the time to stop assuming that small changes will be safe, and
start assuming that they will break something unexpected.  It is a sad
fact that the discovery of important problems takes place year-round,
but Eclipse releases only happen once a year.  A problem discovered in
November is easy to fix; a problem discovered in May is hard.