Solving the Progress Bar Paradox

[breck7]

I am uploading a PDF and waiting for it to finish. But I think I've run into the Progress Bar Paradox

The Progress Bar Paradox If the job is taking a while it will probably fail so I should abort. If the job is taking a while restarting will take even longer so I should not abort. What do I do???

But there's a simple solution that I don't think anyone's tried that could be great. You can have the world's best progress bar. It would show the distribution of expected success and expected wait times all in the same space as a traditional progress bar. A user can see exactly when they should "call it quits" vs "wait it out", because they could see the expected value of waiting longer.

Sketch:

Benefit: Help the user out by showing them the distribution of successes over time, so they can see probability of success for various wait times, and can make an informed decision about when they should give up.

Units: [] Simple server side endpoint to collect job start ids, in progress pings, and job success/error status. This does not have to be self hosted. Something like firebase would work fine. [] Simple script to process the data from the unit above and generate a simple CSV/serialized JSON containing the summary statistics of the jobs above. This does not have to be automated for a long while. You could run it once a month or every few months and save the output as a static. [] Client side progress bar component with a simple API that takes the summary statistics from the unit above and renders a component for the user. My guess for a first prototype would be to have the Y axis be # of successful jobs (y axis max would be the mode) and x axis would be the log(time). [] Theme/Styling/Funny copy for the unit above so people are like "wow, this is the world's best progress bar!" The working name I came up with is "Watermelon Bar", but I'm sure there are better names/designs:

[breck7]

It's happening again :(. Okay maybe soon I'll contribute a PR.

[cindywu]

This is something we should work on. Even I don't have a lot of trust in flubber when uploading PDFs. Sometimes I say prayers to help this work, which is not a way to build technology.