04:33 <ill_logic> isd: Interesting, I'll look into it. Really, I would ideally just change the timeout. I don't want the grain running for no reason.

16:48 <isd> ill_logic: you can drop the lock when the download completes. But I agree, it's a hack.

16:49 <ill_logic> The problem is that, if there's a connectivity problem, I might fail to drop it.

16:50 <isd> Don't know what language you're using for this? In go I'd just use a defer statement right after I acquire it, so it drops on return from the request handler. Could use a finally block in many other languages.

16:57 <ill_logic> Oh right, this is server code.

16:57 <ill_logic> I somehow figured it'd be through some javascript API.

16:58 <ill_logic> Well, what if the user closes their browser mid-download? Then the server is left hanging.

16:58 <ill_logic> I'd need a background process always running. I have to figure that out.

16:58 <ill_logic> Right now I have Flask.

16:58 <ill_logic> There's a request handler.

17:02 <isd> ill_logic: The server should at some point get a disconnect. Just put the drop in a finally: and it should do the trick.

17:18 <ill_logic> To clarfiy: This isn't one request. I'm facilitating the upload of tens of gigabytes.

17:19 <ill_logic> This is many requests.

17:19 <ill_logic> Perhaps it would be beneficial to maintain a wakelock for each chunk (10 megs or so)

17:20 <ill_logic> But what if the problem occurs between requests?

17:21 <ill_logic> So that's why I mention request handler vs background process. To have a wakelock for the whole upload, I'd need a background process that starts on the first chunk and ends after the last chunk, or if I don't hear from the browser for a while.

18:01 <isd> Yeah, you could have a timeout of some variety.

18:01 <isd> Is there a way you could not lose the necessary state when the app gets shut down? future requests should wake it again. What actually needs to be kept?

18:02 <isd> (Just in general we need to make large uploads work better. You are by no means the first person to have struggled with this kind of thing).

18:27 <ill_logic> Hmm. I wonder if I'm repeating the work of others.

18:28 <ill_logic> As far as maintaining state, I mean, yeah the file is still there. My approach now is to have it automatically retry.

18:28 <isd> I know there's been discussion on the mailing list in the past; you might dig through the archives to see if there's any insight to be fine.

18:28 <ill_logic> If need be, I guess I can allow manual resumes.