Erik Stephens wrote:
Is it my fault that Zope people are overly sensitive right now? Am I spreading misinformation? I thought I was just asking questions. I didn't know that was such a crime.
-Erik
For my part, I don't think we're overly sensitive. What we do though is try to fit "lowest common denominator" when we do releases. In fact, our binary distributions of Zope for linux include a Python that is linked against a very old glibc -- it's actually built on a RedHat 5.2 system for compatibility. If you try to use that in production, it will be *painfully* slow. For example, being the speed freak that I am, I build a custom python 2.1.2 for my own use using gcc 3.1 (thats right, bleeding edge from CVS) because it can make me a python that's about 14-18% faster without me having to go license the Intel C compiler. You do have to throw more options at gcc and massage it though, so that build represents a large portion of risk. What we want to do is lower risk, and that usually means distributing things in a fairly plain fashion and leave people to build better solutions for themselves. However, we can easily cause things to spin out of control when we offer advice or suggestions based on partial data or hearsay. Sometimes this is useful -- but sometimes that rather conditional advice gets picked up as gospel and used out of context. Here's another request we get a lot: distribute a python with large file support enabled. We'd love to do this -- but that means having to either have some kind of unified installer which can auto-sense if your platform can support it, or permuting our builds so that we now have 2 times as many combinations as before. Our position is that we want to provide people with a single choice that works as reasonably as we can make it in all circumstances. You as an end user still have the opportunity to package your own solution -- that's why we're open source. -- Matt Kromer Zope Corporation http://www.zope.com/