[IPython-dev] Stand-alone single-cell live public demo

[Kyle Kelley]

Since you said thread and the details seem right Jason, I'm assuming you
meant to respond to the whole list. If not, please banish me to the shadows.

I'm not currently going to recommend running your own tmpnb to back
widgets, since it requires a whole bunch of security constraints for me to
feel it's reasonable to be hosting it. Docker does most of the sandboxing
but then we take additional steps at an operating system level. These are
in Ansible as well, publicly on github but that doesn't make it any easier.

What I'd like to do is provide a much more simpler way of interacting with
kernels that you can point to, as an API. You get the compute on demand, in
a similar to way as seen in the beta.oreilly.com content. In the backend,
that's running on tmpnb too but you as a user shouldn't have to manage that
setup. You also don't necessarily want to pay for the kernels per your
visiting user (maybe up to a point, as tmpnb does). Perhaps it gets
offloaded to another provider of a kernels service. I've still got learning
to do on my end.

Can you provide access to the GUI you've developed to some number of us (or
to everyone on the list)? I'm not in academia so I promise not to scoop
you. ;)

-- Kyle


On Fri, May 8, 2015 at 7:16 AM, Jason Grout <jason-sage at creativetrax.com>
wrote:

> On 5/7/15 19:59, William Stein wrote:
>
>> On Thu, May 7, 2015 at 4:38 PM, Nathan Goldbaum <nathan12343 at gmail.com>
>> wrote:
>>
>>> Hi all,
>>>
>>> I asked about this a few months ago, but I'm curious if the status has
>>> changed in the meantime.
>>>
>>> I'd like to make a GUI that I've developed based on IPython widgets
>>> publically viewable. The end result I'd like is a way to paste a link to
>>> the
>>> demo along with the submission of my paper to the arxiv. Each user would
>>> get
>>> a unique live view of the GUI, although there would be no input from the
>>> user besides GUI interactions and everything will be read-only.  Each
>>> instance would need to be able to see a few hundred megabytes of on-disk
>>> data.
>>>
>>> The last time I asked about this I was told I'd need to whip up something
>>> by-hand.  Are there any services that do this already?  Failing that,
>>> what
>>>
>>
>> Jason Grout wrote something kind of like that, which Andrey Novoseltsev
>> runs:
>>
>>      http://sagecell.sagemath.org/
>>
>> We just moved the hosting to both some machines in Germany and some
>> machines on Google Compute Engine, to increase reliability, etc...
>> The above is all BSD licensed, and provides the ability to nicely
>> embed interactive computation in web sites.        Anyways, thought
>> you might find it relevant.
>>
>
>
> The sage cell server sort of works (worked?) with IPython widgets. There
> are two big issues with how well it works with IPython widgets:
>
> 1. it didn't include all of the necessary styling and libraries that
> IPython widgets depended on (e.g., Twitter Bootstrap, etc.), so sometimes
> they looked weird.  A way to fix this would be to recode javascript
> frontends for the IPython widgets that work in the context of any webpage
> (not just pages that use Bootstrap, for example).  Good news is that we
> want to do this, but it will be a while before it is done.
>
> 2. It worked with an older version of IPython widgets, but I have not been
> checking in the last year to see if any of our changes to IPython widgets
> work well in sage cell server.  And we have been making a lot of changes in
> IPython widgets.
>
> The good news is that (1) you have total control over the webpage, so you
> can include the necessary libraries for the IPython widgets to work, like
> Bootstrap.  However, (2) is still a possible issue.
>
> There are also other projects for providing a backend like this.  For
> example, try.jupyter.org or tmpnb.org host ephemeral IPython notebooks.
> That might be the easiest solution.  To find more about setting up
> something like those, please ask in the IPython help room:
> https://gitter.im/ipython/ipython/help, or Kyle Kelley might reply to
> this thread here.
>
> Thanks,
>
> Jason
>
>
>


-- 
Kyle Kelley (@rgbkrk <https://twitter.com/rgbkrk>; lambdaops.com,
developer.rackspace.com)
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.python.org/pipermail/ipython-dev/attachments/20150508/1251ef46/attachment.html>