Quantcast
Viewing all articles
Browse latest Browse all 7410

Checksum on uploaded files

Hello @jo2012,

I agree with you that calculating the checksum of the source file on the server side can help in the event something in the upload process has gone wrong and the resulting file is corrupted, there's no doubt about it.

However, when you realistically look at the chances of that happening, you'll realise that in actuality they are pretty slim.
The upload is done over TCP so faulty packages will be retransmitted and even if the file does end up being corrupt, the transcoding process will most likely fail.
True, in such cases, comparing the hashes between the original file on the client side with that of the server side is the easiest way to understand the cause of the conversion failure but from our experience, this is quite rare.

In the event the transcoding process completed successfully, it is highly unlikely that the source file is incomplete/corrupted.

Considering the small amount of cases in which the uploaded source file ends up being corrupted, along with the fact the hash calculating operation on big files can get a bit consuming [we do sometimes handle very large source files and of course, process quite a lot of source files in parallel at any given moment], calculating the hash was deemed uncritical.

BTW, for ingestion via the drop folder mechanism, specifically, a checksum is calculated.

I'm also interested in the reason why you're asking about it, have you come across many such cases when ingesting files into the Kaltura server? I'm happy to look into any such issues.


Viewing all articles
Browse latest Browse all 7410

Trending Articles