Java Plug-in and Java Web Start provide support for lazy downloading
of jar files. The algorithm used is largely based on a specific list
of jar files specified by the developer.

JPI Example:

   cache_archive=jar1.jar,jar2.jar,...jarn.jar

When a request is made for an unloaded classfile or resource, these
jar files are searched, in the order specified. If the jar file is
not present in the cache, it is downloaded and then searched. If the
class or resource is present in the last jar file in the list, all
previous jars are downloaded and searched.

This algorithm is OK for jar files that are not yet downloaded from
the codebase. However, it is inefficient for jar files that have already
been downloaded and stored in the cache. Once a jar file is downloaded
we should index the contents for future class file and resource
searches. A better performing solution would be to create a high
performing cache index for downloaded jars and update the index as jar
files are downloaded and stored in the cache. The correct search
algorithm for class files and resources is as follows...

1) Check the cache index for a matching key
2) Check each jar in the archive tag, downloading if necessary

Using the cache index would speed up the search algorithm and prevent
unnecessary loading of jar files.



###@###.### 2004-12-06 18:41:19 GMT
###@###.### 2004-12-06 20:08:33 GMT