One of the advantages of SageMath over Mathematica is SageMath's ability to run a remote notebook without continuous network connection. The Mathematica remote kernel requires continuous connection. It's sad that SageMath is regressing to the level of Mathematica in this respect.
Writing output to disk during a long calculation is normal at the production stage. My problems come from intermediate sized calculations and from the testing stage of big calculations. One wants to to be able to go back to the notebook frequently to check the output. I've investigated cocalc. It requires the overhead of Docker. Also, it uses a modified version of Jupyter. This is apparently fine for allowing interrupted network connections. But the interface is apparently not identical. I'm trying to proselytize for SageMath over Mathematica to members of my center who are already using Jupyter for Python coding (for Machine Learning in high energy physics). Two versions of Jupyter won't fly. Also, they use Machine Learning libraries that are not present in cocalc. They can be installed in the Docker image, but too much IT support is required for that. Running cocalc on one's laptop in addition to the server requires Docker on the laptop. Too much trouble. Daniel On Monday, April 22, 2019 at 1:59:49 PM UTC-4, Nils Bruin wrote: > > On Monday, April 22, 2019 at 5:56:42 AM UTC-7, Daniel Friedan wrote: >> >> The kernel is still running. The problem is not the culling of idle >> notebooks. >> >> When I close the browser tab and then reopen the notebook, the kernel is >> still running but output is lost and output from ongoing calculations does >> not appear. When the ongoing calculations are finished, I can interact >> with the notebook again. >> >> Google tells me that this is a long-standing deficiency of Jupyter. >> > > Ouch, I see. I didn't know that. I've tried and can confirm the behaviour. > > The (deprecated) SageMath notebook behaves properly on reconnecting, as if >> there was no interruption. >> >> This behavior is indispensable to me, so I can't switch to the Jupyter >> notebook until Jupyter behaves properly when reconnecting. >> > > This deficiency in Jupyter is not going to make the resources available to > maintain the sagenotebook, so I'm afraid the reality is going to be we're > going to have to do without it. You could take a look if cocalc does a > better job. It has its own notebook and I think they wrote/adapted their > own ipynb frontend too. > > Otherwise, a work-around would be to structure long-running computations a > little differently to write state to a separate file rather than depending > on logging into the notebook. The kernel doesn't seem to lose state; it's > just output, so once the process has completed (which you could then see > from the file), the notebook is ready for interaction again. > > Another workaround is to not close the browser but instead use something > like VNC or xpra to disconnect/reconnect (remotely) to your running browser. > > It's pretty obvious that different behaviour of Jupyter would be the > better solution, but if the deficiency has been around for a long time and > you don't have definite plans to fix it yourself (I wouldn't know how to), > the reality is that this will likely not change. > -- You received this message because you are subscribed to the Google Groups "sage-support" group. To unsubscribe from this group and stop receiving emails from it, send an email to sage-support+unsubscr...@googlegroups.com. To post to this group, send email to sage-support@googlegroups.com. Visit this group at https://groups.google.com/group/sage-support. For more options, visit https://groups.google.com/d/optout.
