On Wed, Jul 28, 2010 at 11:55 AM, kcrisman <kcris...@gmail.com> wrote: > Dear all, > > I recently got back from useR! 2010, the R user conference. This was > the second time the conference was held in the US, this time at NIST > (a government agency in suburban DC). > > The most important thing for Sage folks is probably the talk I gave: > http://www.sagenb.org/home/pub/2270/ The reception was very positive, > and I hope that will lead to some future collaboration, or at the very > least higher awareness of Sage in the R community. They really liked > interacts - no surprise - and that it was painless to download R > packages, even in the notebook! So good results - especially > considering I was only there for 24 hours. > > However, there were some very intriguing things I gleaned from the > various talks and the overall atmosphere. I'll try to summarize these > below. > > ++ > > First, the community was clearly a community, and a large and healthy > one. There were probably 400-500 attending from all over the world, > from government, industry, health fields, corporations, analytics, and > of course academia, though this last was not even a plurality, I > think. People knew each other, and many were people *not* in the > academy who were connected to R via R user groups (see > http://blog.revolutionanalytics.com/local-r-groups.html), which > apparently are pretty common and for which people actually get > together to learn about R and socialize on a regular basis. > > And on the plus side, long-time users agreed that 10 years ago R was > pretty small, and now it's just been exploding. So I think it is > quite likely "the time is right" for other open source solutions of > the third wave (or whichever wave this is). What else can Sage do to > promote community outside of the Sage Days type participants?
Let's start users groups! For the first year of Sage development, I actually was very active in the Boston (and San Diego) Python user groups, and gave a lot of talks about Sage. Those people were great. Maybe the time is right (and ripe). > > ++ > > Second, there was some very frank discussion of R's shortcomings (as > good as it is) which might be relevant to Sage - sometimes making Sage > look ahead of the curve, other places not. I'll try to categorize > them. > > 1. The corporate perspective. Essentially, many people made the point > (including from Merck, P&G, Ancestry.com, and Facebook) that R is > great, but concerns about tech support, varying stability of packages, > integration with MS Office, etc. yield many companies to have R under > the hood but not in interactions with VPs or other end users. One big > solution to that is that several companies have started up to offer > supported R solutions, either to train new users or to provide > specific commercial support to specific versions of R. They are > apparently doing well! > > I see this as relevant to Sage in a number of ways, but certainly when > it comes to potential engineering users and schools hesitant to/unable > to set up their own Sage servers. Big +1. > > 2. Speed and scalability. These are of course related to the previous > one from the org perspective, but also bring technical challenges. In > particular, several people mentioned in talks that R needs to be much > more scalable to HUGE data sets, performance needs to improve, and R > is slooooow. There were several things in the works with this last > thing - include something like Cython, and others not - and there is > also work on things like using multiple cores *intelligently* (since > it slows some stuff down a lot) and enabling arbitrary size input by > avoiding memory, etc. Performance issues are important in real-time > environments. > > Several people were very interested in Cython when I pointed it out, > and I think we are ahead because Python has some parallelization > stuff, right? But it was clear that this kind of thing is a big issue > in the big picture. Cython kicks aRse speedwise. And at least Python *has* threads -- even with the GIL it can be much better than R for dealing with IO -- at least that's how it appeared from a paper I read on difficulties of using R for some realtime trading. > 3. Don't be too clever. This was less often mentioned, but the sense > was that R is so good for graphics and visualization that sometimes > things are too informative, and post-processing is needed for the end > user. > > I don't think this happens with graphics per se in Sage, but sometimes > it is true that things get clever in Sage too, I suspect. Any > examples of end users being confounded by this? > > ++ > > Finally, there was a good talk by an R blogger about promotion in > general and blogging in particular. See for example > http://www.r-bloggers.com/ > . I think that especially Fredrik, Minh, and Martin have done a good > job with this, but we can do more. In retrospect, the little Sage > stickers at the Joint Meetings were just this sort of thing, though > the speaker talked about Twitter, having guest posts, using animation, > using lots of tags, etc. There is even a "video Rchive" out there. > > +++++++ > > One meta-question is how much all this applies to Sage; in some sense > we have a much more limited potential user base, but in other ways we > have at least as big of one and then some. What is different or the > same about Sage and R from this standpoint? R contains Sage, so Sage has a less limited potential user base. > > I hope this provides much food for thought! > Many, many thanks for posting this. I'm really glad you spoke at the R conference. -- William -- To post to this group, send an email to sage-devel@googlegroups.com To unsubscribe from this group, send an email to sage-devel+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/sage-devel URL: http://www.sagemath.org
