On Sat, Jul 25, 2009 at 5:19 PM, David Kirkby<drkir...@gmail.com> wrote: > > 2009/7/26 William Stein <wst...@gmail.com>: >> >> On Sat, Jul 25, 2009 at 4:57 PM, David Kirkby<drkir...@gmail.com> wrote: >>> >>> 2009/7/25 Robert Bradshaw <rober...@math.washington.edu>: >>> >>>> I've learned to never underestimate the ignorance of computer users. >>>> It's not that they're all unintelligent, but the number of people >>>> that just want to use a computer vastly outnumbers the number of >>>> people that want to learn anything about computers. Often they just >>>> want to spend their time learning about different things (just as I >>>> want to drive a car, but personally have no interest in repairing >>>> one). "Download this file and double-click on it" or "go to this site >>>> and create a username/password" is fine for most people, but as we >>>> learned with 7zip, anything more, even with a clear README.txt, is a >>>> surprisingly high barrier. >>>> >>>> Of course, even now, there are lots of windows users (at least it >>>> seems to be the largest download). >>> >>> I wonder if it would be possible to create a setup.exe which someone >>> downloaded, clicked on the setup.exe, which then installed the vmware >>> software, set up Sage, and *everything* with them doing no more than >>> they would with any other windows application? >> >> No, that would be illegal and a violation of the vmware terms of use. >> However, it would be possible to do something like that using one of >> the open source virtualization solutions (e.g., Virtual Box), or using >> AndLinux. I recently started playing around with Virtual Box again, >> and think it's made a lot of improvements in Version 3.x, and that we >> should consider switching to it from vmware for the sage windows >> virtual machine. > > Perhaps consider doing it with VirtualBox then. > >> Using Sage via virtualization in windows has many drawbacks and also >> many advantages. > > I suspect many of those disadvantages relate to unfamiliarity rather > than real technical issues. But I've never been a great fan of > windows, so I've not tried it.
I think there are many many genuine technical drawbacks, including: * size -- any virtual machine solution is going to be about 3GB, instead of 1.3GB. * filesystem -- the files that a vm version of sage uses will be located on some sort of linux filesystem in the virtual machine. Editing these files using native windows tools is painful at best (and involves ssh'ing the files back and forth). * memory waste -- virtual machines have memory overhead from running another operating system, and also they typically have a hard limit on the memory when the machine is started up. Imagine if math program X on Solaris used *exactly* 350MB any time you used it -- never more, never less -- and just ground to a crawl if you used more memory. That's Sage in a virtual machine. * integration with native environment -- one can't use Python-for-windows libraries and modules alongside Sage in the same python process. E.g., a Python module for doing something with excel or outlook or word can't be combined with Sage. This is a major technical disadvantage. * networking -- I have helped people setup vmware and colinux installs of Sage many times, and the number of times that there are weird show-stopper networking issues is way too high. These should not be underestimated, as they case a lot of pain and frustration. These completely go away with a native port since localhost is trusted (I hope). * speed -- there is overhead which can slow down Sage, especially for older computers that don't have hardware support for virtualization. That said, in some cases Sage on vmware properly configured can be faster than a native windows port for raw computations, because e.g., MPIR assembly optimizations are so good on Linux. There is a substantial speed hit with colinux. But every problem above could be addressed to some degree by enough work and understanding and customization of a virtual machine. That wasn't an option with VMware, since it is closed source. But Virtual Box is open source, as is colinux. Probably the biggest problem by far with virtualization (or colinux) is that nobody but me ever seems to post small virtual machines configured for general use. Gary Zablackis used to do a great job exploring and posting various ways of running Sage under Windows. By the way, Sage does almost work under Cygwin, sort of. Michael Abshoff spent a few days getting it going in May and showed me a running version of Sage then. The hardest problem was some issues with libsingular, which he might have sorted out. > On thing I've been puzzling over is that when we get a Solaris binary, > it would be possible for a package to be built. Sunfreeware is > probably the best known site for Solaris binaries. But the person that > runs it, Steven Christensen, has quite a few connections with Wolfram > Reserach; I don't believe he is an employee, but has his own blog on > their site, he is the administorator of the Mathematica newgroup. It > would be intersting to see how he reacts if asked to produce a package > for a product like Sage. That site is sponsored by Sun according to the upper left corner of http://www.sunfreeware.com/. Sage fits with Sun's core mission way way better than Mathematica and many people at Sun have made that abundantly clear to me. William --~--~---------~--~----~------------~-------~--~----~ To post to this group, send email to sage-devel@googlegroups.com To unsubscribe from this group, send email to sage-devel-unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/sage-devel URLs: http://www.sagemath.org -~----------~----~----~----~------~----~------~--~---
