ChromeChunkStore experimental

[kocoten1992]

I'm writing a plugin to get around chrome limitation (1.8GB in memory before it crash), it was able to download 2.2GB in chrome (theoretically it could go up to 20% of available storage)

https://gist.github.com/kocoten1992/ad3af01bb3b0a257775ae6277081deab

One big caveat, have to build up storage upfront, eg. if torrent have:

8e79d607fce0a41976e4ace19f32d36284fe9cc2 // infoHash
2176931963 //length

Need to create a file name 8e79d607fce0a41976e4ace19f32d36284fe9cc2 and fill it with gibberish data of length 2176931963 (don't know if there are option to this, chrome local file api are real weak)

Then:

var client = new WebTorrent
var opts = {
  store: ChromeChunkStore
}
client.add(torrentId, opts, function (torrent) {
  // despite being weak, it seem to have ability to store much data
  // after done download, go to
  // filesystem:http://example.com/temporary/8e79d607fce0a41976e4ace19f32d36284fe9cc2
  // to get your file
})

[kocoten1992]

update code*

[feross]

Nice, have you tried other options like https://www.npmjs.com/package/idb-chunk-store ?

[kocoten1992]

Hi senpai, I'm testing it right now as you mention it, but why you haven't integrated it into webtorrent yet, is there limitation ? It would take week to try it out, your input is much appreciated

[feross]

Some folks have had success with it. @DiegoRBaquero have you used it? If it's stable enough, I would support it being integrated into webtorrent core.

[kocoten1992]

Idb work pretty nice on firefox, occasionally freeze UI on chrome, think it would better if rewrite use web worker.

One thing bug me, videostream only support MP4, and IndexedDB can't be transform to Blob URL (?), if videostream support more format, this could really be a viable choice.

As for now, with the approach upper, file instanceof File // true we could do a Blob URL, but my code upper have bug that when it done download, it not seeding back (getting request time out ? don't know how to fix).