Sage Math Cloud | Hacker News Reader

Sage Math Cloud

sagemath.blogspot.com

58 pointssdenton415 comments

Since I got involved with Sage years ago, I've always liked William Stein's candid way of describing the project's activities. If the project failed to meet a stated objective, Stein would say, "We failed". No public-relations rhetoric, no glossing over, just the facts.

The basic issue for Sage is that it's gotten too large and complex, and too Linux-platform-dependent, to be a simple install on the majority of student machines. And with each new release, it's become larger and less accessible to many end users, most of whom run Windows on a limited-capacity PC.

Early on, someone created an installation package that ran Sage in a drop-in virtual machine that ran under Windows, but that approach quickly unraveled as Sage became larger and too much of a burden for a small system. This means that, at present, unless a person is willing to dual-boot Linux or run Linux alone, they can't run Sage locally.

Sage on a server has problems of its own -- which server, where, at what cost? But on the positive side, even when running on a single system, Sage behaves like a server-side application, so little needs to be changed for this strategy.

I hope this turns out to make Sage more popular and accessible -- IMHO it's a great mathematical tool, and the project's open-source philosophy only increases my respect.

My Sage tutorial: http://arachnoid.com/sage

> too Linux-platform-dependent,

Can you expand upon this? Not even Unix dependent, but precisely Linux? There's no hope of getting it to work with, say, homebrew on Mac OS X either?

There are mac binaries.

Part of the issue for sage with tying into a lot of academic software is it has to tie into a lot of academic software, and that sort of stuff isn't always made with a lot of concern beyond 'can the graduate student compile it on the graduate student's machine and maybe the local cluster.'

That's typically only true for really obscure things, but not for well-established projects like GAP, Singular, or Scipy. So what part of Sage is in this state? Are people contributing to Sage itself but failing to provide any reasonable infrastructure to support that software?

People don't "contribute" math components to Sage, for the most part. Sage is like Debian or Slackware: it is a distro: a "bag" holding a giant bag of any third party program/library that the maintainers are able to compile and run in the distro bundle; plus a high level shell (akin in function to a Desktop/WM like xfce or unity)

I know that there are some pretty big components that are Sage-only, like the sage-combinat package, and I know the people working on it are packaging it correctly (i.e. no grad-student kind of work). I don't know what proportion of the work goes into such components instead of integrating every other major free math software out there.

I'm just surprised to see Sage getting accusations of grad-student level of maintainership code. Are they really scrambling that much outside of the mature math projects for code?

Yes. USA institutions are happy to pay for commercially licensed software, but refuse to pay non-tenured-profs to write open source software, AMD refuse to assert scientific ethics and send their license dollars to open source projects.

Sage builds and is tested on OS X 10.6 to 10.9, Ubuntu, Debian, Fedora, Suse and Solaris 10. See http://wiki.sagemath.org/SupportedPlatforms

Overview for the technical inclined: http://build.sagemath.org/sage/waterfall

How hard is it to install Sage on a server that is local to a college? Especially if a person knows a fair bit of Linux?

Not difficult at all -- in fact, it almost happens by itself. The problem is not with installation, the problems are security and server loading.

Any reasonably skilled server administrator can install Sage. But setting up user accounts and preventing server abuse are more important:

    sage: N(pi,1e3)
     3.14159265358979323846264338327950288419716939937510582097494459230781640628620899862803482534211706798214808651328230664709384460955058223172535940812848111745028410270193852110555964462294895493038196442881097566593344612847564823378678316527120190914564856692346034861045432664821339360726024914127

    sage: N(pi,1e12)

    (smoke starts pouring out of the server ...)

Wolfram Alpha solves this problem by timing out any transaction after a few seconds. I don't know how difficult that would be to set up in the Sage server.

The sage cell server (used for embedding sage computations in webpages) has good features in this direction. Each cell gets run in its own sandbox, with a server configured timeout.

You can install it via NFS and add a few cute icons for the users. Really easy!

The interesting thing in this write up ifs the difficulty of securing funding to take care of some of these basic usability issues, or hire dedicated developers. A lot of the difficulty is that there are lots of people working on sage, but all of the incentives for the involved researchers and grad students is on writing papers and advancing their own research; there are almost no rewards for improving usability. The job market is brutal, and this kind of computational work simply isn't valued by hiring committees.

If this gets popular and they try to sell to UK customers, I would imagine that there'll be a trademark problem.

Reply on news.ycombinator.com