Monday, December 15, 2003, 12:02:02 PM, Niclas Hedhman wrote: > On Monday 15 December 2003 17:32, Daniel Dekany wrote: >> Sunday, December 14, 2003, 11:52:09 PM, Stephen McConnell wrote: >> > Translating this to a computational defintion - what you saying is that >> > it would be better that this criteria be expressed in metadata (e.g. in >> > the block defintion) as opposed to metainfo (e.g. via an interface or >> > xinfo descriptor). >> > >> > If that's a correct translation - then I agree. >> >> Yes. In the case of Merlin this probably means storing this in the >> bock.xml. > > Isn't this contradictory to what you are saying elsewhere?
I don't see contradictory here... unless, you mean that I said that the code that uses the component should not be changed because of implementation changes... and that we have to choose between release always or never, etc. I have pointed out later in that mail, that the solution I have found is a *compromise*, and it does break the "release always or never", and changes it to "never release, except if a Role is explicitly stated to be 'manually released', which should be avoided.". > Do you see the "Release Requirement" tied to the Specification or the > Implementation? "Role" is tied to the Specification. Then I think it should be tied to the Specification. (See the other mail where I explain why.) -- Best regards, Daniel Dekany --------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
