J2SE 1.4 beta install on guilden ...

John Shott shott at snf.stanford.edu
Thu Jun 7 12:08:24 PDT 2001


Bill and Mike:

Because of disk space shortages on guilden, I installed the 1.4beta version of
j2se in /app/j2se ... instead of /usr/j2se.  I have then made a sybolic link
that points from /usr/j2se to /app/j2se.  It is my hope that this means that
all of our Makefiles, install, and run scripts will work normally ... without
any fiddling with those changes.

Here is what I propose to do ... all on guilden:

1. I'll create a new directory ~/labnet1.4beta and then do a fresh "cvs
checkout -P labnet" there.

2. I'll make the standard changes to make it work under the idlj instead of
the idltojava compiler.  I will also need to modify the makefile associated
with the idl stuff so that it uses the older BaseImpl approach rather than the
POA approach.

3. If this works, I'll hope to be able to compile and install the servers ...
and see if our old CORBA problems still occur.

4. If we don't have problems ... great!!!

If we still have problems we will need to strategize ...

One option is going from the Sun ORB to JacORB.  Bill, I've downloaded (but
not installed ...) the latest version of JacORB.  I've also left you a copy of
the JacORB 1.3 programming guide.

Without being "Mr. ORB", I can see that it offers a few features that we don't
seem to have with the Sun ORB.  Among them, it offers 3 different levels of
logging ... as near as I can tell, the Sun ORB doesn't do any logging but only
throws exceptions.  JacORB also allows significant amounts of configuration
including the number of retries, the time interval between retries, etc.

Bill, you'll probably have a lot more to say after reading the JacORB stuff.

One thing I notice, however, is that JacORB seems to only support the newer
POA approach.  I think that means that we can't instantly test out JacORB
without knowing a bit more (at least for me ...) about what is involved from
converting from an idltojava converter that supports POA vs. BaseImpl. 
Clearly some of the files are named different things ... but I don't know
whether the changes are as mechanical as from going from idltojava to idlj
(e.g. changing _ClassOperations to ClassOperations and _ClassTie to
Class_Tie).

Bill, you probably have a lot to add on the subject based on what you've
learned at JavaOne ... I know you talked to the fellow who really liked JacORB
... and I vaguely recall that this evening was the evening that you were going
to be able to confront the Sun CORBA team.

Talk to you later,

John



More information about the coral mailing list