hivemind-commits mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
Subject [Jakarta HiveMind Wiki] Updated: NotXMLProposal
Date Tue, 27 Apr 2004 14:46:10 GMT
   Date: 2004-04-27T07:46:10
   Editor: HowardLewisShip <>
   Wiki: Jakarta HiveMind Wiki
   Page: NotXMLProposal

   no comment

Change Log:

@@ -154,3 +154,18 @@
 HarishKrishnaswamy: Christian, I am not sure how you can construct the service from with
the configuration file using simple closures. Could you elaborate?
 Howard, Beanshell is simply plain Java with some syntax sugar, so it is not necessary that
every line should be an executable statement. I think the use of a scripting language is two-fold
- make the config file smaller and simpler, and also simplify the framework internals. The
SDL, although looks better than XML, still requires a lot of typing.
+HowardLewisShip: I think we can optimize the format somewhat; the SDL version is isomorphic
to the XML version, but even the XML version could be made more succinct. We seem to be in
a situation of ''declarative'' (SDL/XML) vs. ''procedural'' (scripting). To me, scripting
opens up a whole can of worms ... and sacrifices one of the '''key''' features of !HiveMind:
line precise error reporting. Groovy builders are a step up from pure BeanShell statements
but still miss the mark. I think HiveDoc will also suffer. I have nothing against something
+. . .  invoke-factory (
+  script <<
+return new MyServiceImpl(. . .);
+} . . .
+The {{{<< . . . >>}}} syntax is something I'm looking to add into my SDL prototype.
I like the idea of scripting, but I like line precise error reporting much, much more. I also
think scripting introduces further problems if and when there is tool support (the tool has
to parse the script, which is not nearly so regular as the SDL or XML).

To unsubscribe, e-mail:
For additional commands, e-mail:

View raw message