[IPython-dev] SockJS

[Jason Grout]

On 6/26/12 5:18 PM, Fernando Perez wrote:
> On Tue, Jun 26, 2012 at 2:51 PM, Brian Granger<ellisonbg at gmail.com>  wrote:
>> We are aware of these alternatives.  My philosophy (and I think it is
>> reflected in the rest of the team) is this:  by sticking with plain
>> old WebSockets, all we have to do is *wait* for the situation to
>> improve.  That is super easy to do!  Once IE 10 comes out, we should
>> cover most browsers.  If we go another route, *we* have to do all the
>> work, and in the end, we will throw that work away because WebSockets
>> will eventually have universal support.  That sounds like extremely
>> low priority work to me.
>
> +1 (and also with Min).
>
>> People should not be running older browsers anyways.  There is simply
>> no reason, you almost have to try hard to run an older browser.
>
> There are real-world scenarios where individual users are *forced* to
> run older browser by constraints beyond their control, such as
> university/school/company-run machines where they have no admin
> control at all.  So I do see that *in principle* it would indeed be
> nice to support older browsers.
>
> But given the extremely high cost of development to us of such
> support,

I thought SockJS was pretty much a drop-in replacement for Websockets. 
Theoretically.  Of course, we haven't actually tried it out.

Or I suppose you're also talking about lots of other quirks with older 
browsers beyond this issue.


> I just don't think it's a realistic proposition: we'd get
> bogged in a tarpit of old browser support that would easily consume
> all our limited dev resources.
>
> So I simply accept that IPython isn't the right tool for those in such
> constrained environments, unfortunately.  We have to pick our battles
> carefully if we're going to make progress, and this is one such
> tactical choice.

Thanks; that sounds like nearly unanimous replies from the three major 
people.  We were in much the same position a year ago regarding IPython; 
we could wait for IPython 0.12/0.13 to mature to base our sage cell 
server work on, or we could forge ahead with the Sage cell server on the 
old IPython.  I'm glad we forged ahead; we've learned a lot about how 
the environment works, and we've had a lot of adoption.  Now we're 
rewriting it all to use the IPython framework this summer :).

I see more clearly the different audiences we cater.  IPython can force 
a browser choice because the using the notebook is usually a personal 
choice for a personal user.  In our case, universities are a primary 
target, so we'll use SockJS in our deployment of IPython/Sage web 
interfaces, which will hopefully go away in a few years when websockets 
is ubiquitous.  The nice thing is that (theoretically) SockJS is a 
drop-in replacement for websockets, so it should be really easy to push 
things back to websockets when the time comes.

Anyways, good to know it was considered!

Thanks,

Jason