|
Size: 8676
Comment: cleaned up
|
Size: 8677
Comment:
|
| Deletions are marked like this. | Additions are marked like this. |
| Line 9: | Line 9: |
| Here's a conversation that news of Google Chrome OS triggered, where William Stein outlines his reasons for strongly preferring developing a web-based interface to Sage, rather than stand-alone graphical interfaces. | Here's a conversation that news of Google Chrome OS triggered, where William Stein outlines his reasons for strongly preferring developing a web-based interface for Sage, rather than stand-alone graphical interfaces. |
Ahmed Fasih Aldebran
Email: <fasih.1 AT SPAMFREE osu DOT edu>
Sage stand-alone GUI versus web-based Notebook
Here's a conversation that news of Google Chrome OS triggered, where William Stein outlines his reasons for strongly preferring developing a web-based interface for Sage, rather than stand-alone graphical interfaces.
<aldebrn> I was somewhat put off by this line in the Google Chrome OS release: 'For application developers, the web is the platform' but then I remembered the intense Sage Notebook experience and now am optimistic
<wstein> the web is the platform -- drink the coolaide
<wstein> I'm excited about Google Chrome OS, since I'm curious to see how Google compares to Apple at taking a UNIX OS and making it mainstream.
<aldebrn> Something Alan Kay said influenced my attitude towards webapps: in the 50s, it was a big breakthrough to realize that one could ship code with data that knew how to work with it. The markup languages of the 80s to today represent a huge regression because we're back to shipping just data without any code, instead relying on the recipient's browser to interface with it
<Dcoder> amen
<aldebrn> Though I guess a modern day answer, Java VM, isn't that great (for various reasons)
<wstein> I personally like web apps. I don't know if they are a regression or not.
<wstein> I just know there are a *lot* of "everyday people" out there who don't know anything about computers, but really like
<wstein> using webapps too. A big plus is that collaboration and community features are much easier to implement.
<wstein> When I started using computers a lot -- in the 80's it was all PCs (atari, commodore). Then things got a lot better
<wstein> and exciting when I went to university and got to to use VAX and U
<wstein> Solaris servers + gopher, archie, etc.
<aldebrn> Of course. I just wonder if Alan Kay's hunch was right, maybe the web would be a whole lot cooler if somehow data was shipped with custom code that interacted with the data, much more than JavaScript/CSS maybe
<wstein> The software looked a lot less pretty (plain text screens), but was way more social.
<aldebrn> The Notebook is amazing, and I really wonder how much effort went into making it work as well as it did
<Dcoder> i dislike webapps and the whole cloud sense in the sense that they relinquish the control of what one runs back to whoever owns the site
<wstein> Not entirely.
<wstein> The sage notebook is a webapp, but you can run it 100% locally.
<wstein> Of course, there are very few examples of this.
<wstein> The vast majority of webapps have exactly the problem you point out.
<wstein> I would like gmail a lot more if it had a 100% offline open source version that used IMAP (or something) to synchronize with a server.
<wstein> Oh well.
<aldebrn> Would it have been easier (time-wise, maintenance-wise) to write an internet-aware stand-alone Notebook that interfaced with a local/remote Sage, and did all kinds of cool display/visualization/programming things than relying on IE/Firefox/Chrome/Safari?
<aldebrn> Stand-alone as in, an executable app that you downloaded and connected to sagenb.org or localhost:8000
<wstein> aldebrn -- definitely not.
<wstein> It would have been vastly harder to write a standalone app.
<aldebrn> Ok I'll take your word for it
<wstein> I've put a grand total of just over 6 weeks fulltime work on the notebook, since it started.
<wstein> And I did most of the work.
<wstein> Try developing that sophisticated of a GUI app that has a native look on OS X, Linux, and Windows, and interfaces with a network server in that
<wstein> amount of time. I think it is not possible.
<aldebrn> Incredible, and that's inclusive of all the great tab-completion, AJAX action, etc.?
<wstein> At least it isn't for me, and since I did the work, that says something.
<wstein> Yes.
<wstein> The Sage notebook is a web app simply because it was by far the easiest thing to write. (At least for us developers.)
<wstein> And I do have a lot of experience with writing traditional GUIs. That was part of why I didn't do it again.
<wstein> Probalby the closest thing to the Sage notebook for math software using a traditional GUI is
<wstein> wxmaxima: http://wxmaxima.sourceforge.net/wiki/index.php/Screenshots
<wstein> Also, Google funded a Google Summer of Code in 2005 to write an "Ipython notebook" as a traditional GUI.
<wstein> Two studnets + 1 enthougt mentor worked all summer on it, and the result was a bad broken notebook-like thing that only worked on OS X.
<wstein> Robert Kern gave a talk about this at Sage Days 1. It really scared me off from taking that approach.
<wstein> There were many other people who did a lot of work on the Sage notebook -- and I don't think I clearly indicated that at all above.
<wstein> So just let me say that though i wrote a lot of it, so did many other people (e.g., Alex Clemesha, Tom Boothby, etc.).
<aldebrn> No I understand. That's a very valuable piece of insight, it should be somewhere on the wiki
<wstein> But the total time wasn't very much.
<wstein> I'm not at all opposed to having a traditional GUI interface. It's just that it's too hard to do given limited time.
<aldebrn> I thought that getting it all to work well in IE/Firefox/etc. would take much longer than writing a GUI interface
<wstein> Well it takes time, but it is *easy to parallelize*.
<wstein> E.g., for getting the notebook to work with IE, I just asked Tom Boothby to work on it, and he spent one week and did it (for pay).
<wstein> wxmaxima seems to have been improved a bit since when I looked at it 3 years ago (when the sage notebook project started); I'll have to try out the new version..
<aldebrn> Especially when I started using the HTML cells in the Notebook to annotate my code
<wstein> aldebrn -- last summer at ISSAC the directory of development at Maple gave a talk about the new features in Maple 11.
<wstein> One of the "major new features" was that tables could have colored backgrounds.
<wstein> That's the sort of thing that just comes for free to the sage notebook.
<wstein> aldebrn -- we should really have a mode in the notebook where it catches all NameError exceptions and auto-defines the variables, then reruns the code.
<aldebrn> So skills in webapp development is definitely important, ok, I'll brush up my JS/Ajax. If Sage Notebook had zoom-pan-trace features for the plots it generated, it'd be just about as useable as the Matlab GUI
<wstein> aldebrn -- we could add such plotting very nicely with Geogebra or maybe FLOT.
<wstein> Geogebra wold be the java way, and FLOT the javascript way.
<wstein> wxmaxima has improved over the years -- I just tried it out.
<wstein> If somebody were serious about making a local GUI-based sage notebook, I would suggest they start with improving wxmaxima
<wstein> to work with more than just maxima....
<wstein> (for the record)
aldebrn was thinking about some incredibly lightweight GUI tools like Shoes: http://shoooes.net/
<aldebrn> I also come from the numerical world where large-scale data visualization is normally not done in web languages. I am content to wait and see what Sage throws up regarding this; given the pace of development, I won't have to wait long
<wstein> My impression is that large-scale data visualization in the numerical world is already done by Sage users using VTK/mayavi and also some Enthought tools.
<wstein> And matplotlib for 2d.
<wstein> But doing something like that over the web is a good challenge!
<wstein> Right now there is a lot of discussion on the SPD list (basically numerical version of sage) about doing offscreen rendering
<wstein> using vtk of 3d images then displaying them in the sage notebook.
<wstein> This is mainly for applications to fluid dynamics (I think).
<aldebrn> Our radar datsets are gigabytes large and we form 10K by 10K pixel frames that we zoom in and out while playing
<aldebrn> It would be great to replace the propreitary software we use now with a webapp
<wstein> It's got to be possible. Google maps is super good at visualizing similarly complicated data (the whole world)!
<wstein> It would certainly be possible to have something like Google maps for visualizing 2d-ish data in the sage notebook.
(2009-07-08, after 1030 EST.)