On Dec 22, 2008, at 8:53 AM, Holger Hoffstätte wrote:
Michael McGrady wrote:
No problem with them being bad. I agree.
huh? Michael, no offense but you keep contradicting yourself. :(
I am not trying to lay down a final solution here but to discuss
issues. I was looking at one and let the other slip out the backdoor,
somewhat like the proverbial picture of keeping balloons under water.
So, I was trying to emphasize the need to have JavaSpaces separate and
dropped the ball on realizing that JINI needed Entry for uses other
than JavaSpaces. Glad Niclas mentioned the error. Not really
inconsistent unless what I put down was taken as written in stone -
maybe soapstone? <g>
The problem is that there is no natural owner for the generic
interfaces
other than Java itself, so far as I can tell.
You seem to associate WAY too much inherent semantics with package
names.
They are not behaviourally binding, and it distracts from the much
more
important constraint of *physical* coupling, i.e. the jar file.
Yes. You are probably right. I do so with a view not only to the
physical state, thinking these divisions normally end up as JAR files,
but also to the ease of understanding the system for everyone
involved. The more important consideration, I agree, is the
"physical" coupling.
So a
simple decoupling into -api jars is all that's needed, and from what I
could tell that is exactly what Niclas has been doing in his branch.
If the only issue were physical decoupling, then this would be
sufficient. I am not at all convinced that package structures
(naming) are as lacking in importance as you indicate. Certainly if
we get a physical separation that would be a huge step forward.
Mike
