We provide a popular open source Java FTP library called edtFTPj.
We would like to drop support for JRE 1.3 - this would clean up the code base and also allow us to more easily use JRE 1.4 features (without resorting to reflection etc). The JRE 1.3 is over 7 years old now!
Is anyone still using JRE 1.3 out there? Is anyone aware of any surveys that give an idea of what percentage of users are still using 1.3?
Why not have your program report back what version of Java it is being run with. This will give you an idea of your user base.
I highly recommend dropping support for Java 1.3, and instead of doing a minor upgrade to Java 1.4, why not use Java 1.6? There have been massive improvements since 1.3. You really are missing out.
Even JDK 1.4 reached the end of its support life in Oct 2008. I think you're safe.
But don't take it from me. The people that you really need to ask are your customers. Maybe putting a survey up on your download page and soliciting feedback will help. If no one asks in three months, drop it.
It's difficult to give figures for what companies use internally. There do exist figures for browser plugin installation, but Sun's figures are confidential.
1.3 support stopped some time ago (Solaris 8 vintage support dragged on for a bit). 1.4 has completed its End of Service Life, but is likely to be supported under Java for Business for yonks. IIRC, if you try to download 1.4 from the archives at sun.com then you are asked for some information such as an e-mail address. 1.5 is more than half way through its service life (but its still quite common on Macs).
It's not that difficult to use 1.4 features optionally. You just need to load one class via reflection (or just package it differently) and then have a 1.4 and 1.3 implementation of a light abstraction over the new features.
Sun allows you to buy support packages for depreciated software such as JRE 1.4. For banks and some other organizations, paying $100,000 per year for support of an outdated product is cheaper than upgrading. I would suggest only offering paid support for JRE 1.3. If anyone needs support for this, they can pay for a hefty support package. You would then shelve your current 1.3 code base, and if a customer with a support contract requires a bug fix, then you could fix the 1.3 version for them, which would likely just mean selectively applying a patch from a more recent version.
End-of-life is normal part of software's life cycle.
The real question you should ask/answer is whether you have a compelling business need to add features to the "old" versions. If not, you can continue to offer it for customers who need it -- but encourage everyone else to take the latest & greatest which requires 1.4 (or 1.5/1.6).
Why not only do critical patches for 1.3 & 1.4 if anyone actually requests them and do all new releases on 1.5 - the current oldest version supported by Sun?
I was developing with jdk 1.4 for a long time while jdk 6.0 was out. We couldn't upgrade(clients this, servers blablabah). At some point, we just upgraded without talking too much about it. Clients upgraded without being annoying, "this upgrade will fix lots of security holes, many bug fixes, improved performance :-)".
Right now, I try to keep my code compatible with jdk 1.5, I have no concern at all for people running 1.4 and below. At some point, they'll understand that it is in their best interest to "try" upgrading.