Thursday, October 09, 2008
We have been running a poll for the past couple of weeks over at LispForum asking people to tell us which Common Lisp implementations they use most. I don't think it would surprise anybody that SBCL is leading the poll, but it might surprise people as to the magnitude. Currently, with something like 97 votes in, SBCL has 53% of the votes. The next most popular implementation is CLISP with about 18%. This is followed by a tie between ClozureCL (OpenMCL) and Lispworks.
My takeaways here are the following:
- First, this poll has limited sample size. 97 respondents, all of whom are LispForum participants isn't necessarily representative of anything. LispForum has a lot of newbies (and indeed was created to help foster that), so it could be that all the old, crufty smug Lisp weenies hanging out on comp.lang.lisp have a completely different usage profile. Still, it's more data than I have seen anywhere else. Perhaps some of those c.l.lisp residents will register on LispForum and vote.
- My biggest takeaway is that SBCL has really moved beyond CMUCL. My hunch is that having a system that is relatively easy to build and is released on a regular monthly tempo is simply a superior value proposition. William Newman nailed it when he forked SBCL from CMUCL and created a healthy, vibrant community around it. I don't dare say that CMUCL is "dead," but clearly SBCL has taken that ball and run with it. CMUCL has only 1% usage in the poll.
- CLISP seems to be a common "backup implementation" for environments where SBCL doesn't run well. For instance, my own personal preference is SBCL on Linux and CLISP on Windows, at least for the time being. I'm also using ECL for a project I'm working on right now, primarily because it has a small memory footprint and low performance requirements.
- Lispworks is certainly leading the commercial implementations. The low price for a high-quality implementation is probably the big draw there. IMO, if I wasn't using any open source version of CL, I'd probably be using Lispworks.
I think SBCL's eclipsing of CMUCL happened several years ago, really...
Undoubtedly. Nobody ever quantified it, though.
I think the poll would have been more worthwhile if SBCL had been split into SBCL/Linux and SBCL/Win32.
Post a Comment
Links to this post: