river-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Holger Hoffstätte <hol...@wizards.de>
Subject Re: Jini, JavaSpaces, JEE, some questions, and some other development issues and ideas.
Date Tue, 02 Sep 2008 10:35:19 GMT

Wade,

In order to understand the deeper and sometimes counter-"intuitive" (which
I often find to be negative conditioning) reasons why things in Jini are
the way they are, I found it very helpful to go straight to the sources,
researching back all the way to 1998. I know this is not a popular
approach anymore, but in this case it is helpful and saves from having to
poke around in idea-space "guessing" for motivations and reasons which
only lead to stray conclusions.

A particularly helpful resource for this is an interview series at Atima
called "A Conversation with Ken Arnold" which can be found on
http://www.artima.com/intv/. Part 4 (Sway with JavaSpaces) has an in-depth
explanation and should help understand the reasons for this approach.

You are not *supposed* to have that additional logic on what you call
"transfer objects"; the core issue does not revolve around behaviour, but
*identity* - one of the three defining aspects of objects as defined by OO
(state, behaviour, identity). For distribution in scenarios with partial
failure (not assuming a closed-world / single-system image [1]), you
*must* let go of the identity. Once you understand that, you will realize
that the behavioural aspect becomes a potential problem instead of a
desirable property, and that is why it is missing from Entries.

Different may not imply better, but better always mandates different. :)

Holger

[1] I wanted to add a looong paragraph about "transparent"
identity-perserving clustering ala Terracotta but that would go way off-topic.

Mime
View raw message