Comments on: Something to watch: Ruleby, a Ruby RETE Implementation http://www.pathf.com/blogs/2007/12/something-to-wa/ Running commentary about agile development, user experience design and Ajax. Fri, 05 Dec 2008 18:16:24 +0000 http://wordpress.org/?v=2.5.1 By: peter lin http://www.pathf.com/blogs/2007/12/something-to-wa/#comment-919 peter lin Thu, 31 Jan 2008 17:48:06 +0000 http://www.pathf.com/blogs/2007/12/something-to-wa/#comment-919 <p>There are some major design and performance considerations for Ruleby to become a first class RETE implementation. The first is the issue of duck typing, which is a complex issue. Even though I've suggested some potential solutions to solve duck typing issues, it's partial at best.</p> <p>To really handle duck typing correctly and still maintain good performance with Ruby is going to be very challenging. The second issue which may take 3-10 years to solve is Ruby's performance.</p> <p>by far the biggest issue facing ruleby is how to handle dynamic modification of objects at runtime. If you allow it, the RETE network topology will take a huge hit and it can't use object type nodes. Instead, it has to take the older relational design Dr. forgy described in his thesis. But even that doesn't begin to address the issue of what happens when an object adds and removes attributes at runtime.</p> <p>My best guess is that a proper and mature implementation for Ruby is going to be atleast 10 years off at best. that's assuming the ruleby guys can get a solid understanding of RETE algorithm within 2 years. I tried to help them out last year and gave them suggestions, but learning RETE takes years.</p> There are some major design and performance considerations for Ruleby to become a first class RETE implementation. The first is the issue of duck typing, which is a complex issue. Even though I’ve suggested some potential solutions to solve duck typing issues, it’s partial at best.

To really handle duck typing correctly and still maintain good performance with Ruby is going to be very challenging. The second issue which may take 3-10 years to solve is Ruby’s performance.

by far the biggest issue facing ruleby is how to handle dynamic modification of objects at runtime. If you allow it, the RETE network topology will take a huge hit and it can’t use object type nodes. Instead, it has to take the older relational design Dr. forgy described in his thesis. But even that doesn’t begin to address the issue of what happens when an object adds and removes attributes at runtime.

My best guess is that a proper and mature implementation for Ruby is going to be atleast 10 years off at best. that’s assuming the ruleby guys can get a solid understanding of RETE algorithm within 2 years. I tried to help them out last year and gave them suggestions, but learning RETE takes years.

]]>