[IPython-dev] Live sync as a service
ellisonbg at gmail.com
Wed Apr 30 22:43:50 EDT 2014
> Why do you have to, given that I didn't implement OT for IPython sync
> for SageMathCloud?
Ahh, I was under the impression that you did. Thanks for clarifying this point.
> The requirement for the differential transforms approach is that you
> have a way to compute diffs and apply patches, subject to some axioms.
> That's pretty much it.
> Anyway, using differential transforms is more or less like using git
> (with commits every second of activity), so conceptually it feels
> familiar. The difference is that instead of getting merge conflicts,
> the patch application algorithm is by definition "best effort" -- you
> apply as much of the patch as you can, and that's it. Things then
> converge quickly.
This sounds simpler than OT for sure. When we get to tackling for for
real in IPython (still a ways off) we will have to dig into this.
> When I looked at
> earlier today I noticed it makes the possibly dubious claim that "OT
> is the technology behind awesome products like Google Docs and
> Etherpad". When I was researching sync approaches last year, I came
> to the conclusion that Etherpad (which is open source now, I think)
> uses OT, but that Google Docs does not. More precisely, Google Docs
> and Google Wave *did* use the OT approach, but then Docs was switched
> to differential sync (Docs was some non-Google product that Google
> bought and rewrote over many years...). That said, I wasn't able
> to be 100% sure, since I didn't confirm this with anybody working on
> Docs at Google, except that the main person (Neil Fraser) behind
> Differential Sync works at Google.
> I'm planning to do more work on simplifying/rewriting/streamlining my
> sync code in the next few days, which will make fixing some bugs
> easier, and also make it easier to make an open source single-project
> version of SMC (which will still have IPython in it, of course).
It would be great if the real time sync parts of this evolve in a way
that can be reused elsewhere.
Thanks for the comments!
> -- William
>> On Wed, Apr 30, 2014 at 10:32 AM, Thomas Kluyver <takowl at gmail.com> wrote:
>>> I just saw that this company is offering live sync as part of their
>>> platform, using operational transforms, which from what William Stein told
>>> us is the harder but more powerful approach to the problem:
>>> I doubt that we want our live sync to rely on a third party service, but it
>>> might be worth looking at their APIs for ideas.
>>> IPython-dev mailing list
>>> IPython-dev at scipy.org
>> Brian E. Granger
>> Cal Poly State University, San Luis Obispo
>> bgranger at calpoly.edu and ellisonbg at gmail.com
>> IPython-dev mailing list
>> IPython-dev at scipy.org
> William Stein
> Professor of Mathematics
> University of Washington
> IPython-dev mailing list
> IPython-dev at scipy.org
Brian E. Granger
Cal Poly State University, San Luis Obispo
bgranger at calpoly.edu and ellisonbg at gmail.com
More information about the IPython-dev