CODEWITHSUNDEEP
apachefirefoxvideogecko
Reut Sharabani
Reut Sharabani

Reputation: 31349

Understanding video files http range requests in Firefox Gecko

I need to download a part of a video file (vp8/webm, can add more data if needed) based on the start and end time inside the file.

I use the requests library to make the requests, and I have an Apache server set-up that allows me to fetch parts of a file using http-range requests.

When I use a player in the browser, it somehow (and that's the part I'm trying to figure out) manages to estimate the entire video duration. This is important since it enables the player to know the offset in bytes from which to make a range request.

So, if there are 1000 bytes on a 1-minute file and I want seconds 30-40, it will make at least two range requests:

A dummy using a header (aborted immediately):

First Request

Now the response contains content-length in bytes (and nothing more that seems informative other than content-type: video/webm).

Now I assume it somehow knows:

total length (bytes) - 1000
total_duration (seconds) - 60
start_offset (seconds) - 30
end_offset`(seconds) - 40

What I think it does next is something like:

((start-offset / total_duration) * total_length)-((end_offset / total_duration) * total_length) = ... = 160

And now it has the range needed to stream the specific part of the file:

Second request

My question is: How an Firefox tell the total duration of the file only by knowing it's size in bytes? When I try and figure out the bitrate avconv gives N/A.

avconv output on the remote file:

  Duration: 03:50:15.26, start: 0.000000, bitrate: N/A
    Stream #0.0: Video: vp8, yuv420p, 352x240, PAR 10:11 DAR 4:3, 15 fps, 15 tbr, 1k tbn (default)
    Stream #0.1(eng): Audio: vorbis, 48000 Hz, stereo, fltp (default)

After inspecting multiple requests made from the browser I am almost certain the video bit-rate is not constant.

I base my understanding around this post. It may be outdated / I may have overlooked something.

I've also looked at StackOverflow posts like this one, sadly they weren't in-depth enough for me to understand how to do this.

Upvotes: 0

Views: 1096

Answers (1)

szatmary
szatmary

Reputation: 31160

You assumption is wrong. The first request starts to download the file starting at byte 0. The browser aborts the download once the file header is received. This header contains the information needed to determine the file duration, as well as 'time to sample' and 'sample to offset' data. Using this information the browser knows exactly what offset to request for a specific time.

Upvotes: 1

Related Questions