[WDSCI-L] Re: EAR/WAR Deployment

["Joe Pluta" <joepluta@xxxxxxxxxxxxxxxxx>]

> From: Mark Phippard
>
> But isn't it just dependent on your definition of the root location?
>
> For example, I would consider the root for an installed WAS 4 AE
> application to be this:
>
> /QIBM/UserData/WebAsAdv4/instance_name/installedApps
>
> And for WAS Express to be:
>
> /QIBM/UserData/WebASE/ASE5/instance_name/installedApps/cell_name

But while you (might) have control over the instance name, you don't have
any control over the cell name, since it's different for every machine.

What part of this are you missing?  When I go to deploy this on a machine, I
have to know the cell name.  That cell name is different on every machine.
And so, while with WAS35, and even Tomcat, I can use the exact same naming
convention on every machine, with this new naming, EVERY MACHINE WILL HAVE A
UNIQUE DIRECTORY STRUCTURE.  In order for me to know where to copy a file, I
have to know that entire path name, and it will be different on every
machine.

I would think with your experience in application deployment that you'd see
the incredible burden of the new technique.  If I am a software vendor, for
example, I have to somehow get the cell name into my deployment process.
This is not a good thing, since it's a manual step and thus prone to user
error.  And it's unnecessary, as the more flexible design of WAS35 shows.

In any event, if you can't see the problem, then I'm shouting at a brick
wall.  This horse is now officially dead in my book.

Joe

P.S. I have identified one major problem with symbolic links.  Neither ZIP
nor JAR can handle them.  If you have a symbolic link, rather than just
include the link in the ZIP file, it attempts to include what the link was
pointing to.  In my case, since I have a recursive link, the utilities fail
miserably.