On Sat, Jan 18, 2003 at 03:50:07PM +0530, Abhir Joshi wrote:
On Sat, Jan 18, 2003 at 12:14:36AM +0530, Manish Jethani wrote:
Well, actually, having a Java interpreter would be a lot cleaner than writing all your s/w in C/C++. Programs like Gautam's are better written (and maintained!) in Java. That's why Java (or Perl, or some other high-level language) exists!
Gautam's program is something with very basic functionality; it is like an enhanced fetchmail (I haven't tried JFetch yet. I downloaded it but I don't yet have JVM). Many more people would accept it if it could be converted into a standalone binary.
There was a specific reason that I chose to use Java in this case. JFetch allows you to easily plugin mail-filters which is really kewl. So all it takes is creating a filter complying to certain interfaces, and dropping it in the classpath. Configuration etc. all can be provided through one central config file. Java's dynamic class-loading mechanism helped me a lot here. As I said earlier, I would rather have this flexibility than worry about a running JVM.
JVM's have become faster and faster over the years and anyone following the Java world would know about technologies like Hotspot which are dramatic improvements in the JVM implementation.
Since JFetch is an enhancement to fetchmail and not radically different from fetchmail, I feel that instead of writing a totally new software, a patch to fetchmail would be more helpful. ESR should be told about this (It's possible that Gautam has already had a correspondence with ESR). If he refuses to include this functionality for some reason, then a new software is worth. Having the functionality added to fetchmail would also mean that users would just have to upgrade their existing fetchmail and add the new options to the existing fetchmail configuration file.
I use a dialup the amount of time I spend online needs to be a minimum. I used to be a fetchmail user, and I wanted specific functionality like:
- message-id filtering (duplicate message removal) - size based filtering - spam filtering
Anyone who knows the POP3 protocol know that these (and more) filters can be achieved without actually downloading the message. Just the headers are sufficient. Now, fetchmail could probably include the size feature... but thats just one of the many that I wanted. SPAM control was another of my requirements (sender based etc.) And you know about duplicate messages due to people cross-posting. Getting all this into fetchmail seemed unrealistic... though I never got in touch with ESR.
Ultimately it is Gautam's choice. Is there any way to convert the java byte code to native code binary? If there is, then those binaries could be distributed, and people like me, who don't have JVM, could start using it right away.
Bytecode->native converters exist today... but FAIK they are quite unstable.
BTW, can anyone shed some light on how plugins are/can be implemented in C/C++ (like GAIM etc.) Any pointers will be greatly appreciated.
-- Abhir Joshi http://education.vsnl.com/abhir/
Cheers, -g