screen recordings sound pointless to me. either it builds reproducible or it doesn't. videos change nothing, videos proof nothing. ASCII replays proof even less 😅
Has anybody noticed that we now have "screen recordings" in our reproducibility tests? As another project is sharing "video proof" of reproducibility, we were asked to also do so but it felt kind of pointless to produce GBs of data for every reproducibility test. We did however start playing around with console recordings that are somewhat more optimized as they record the ASCII on the screen and not every pixel. Resulting files are much more manageable but for example, running the compile script for the Electrum for Android app resulted in 72MB of output. As we test a lot, this is a lot to add in a single day.
Does anybody care about screen recordings? Can we throw them at some nostr relay instead of our git repo, with some expiry date in three months, so that interested users can grab it while it's hot? Any other ideas?
Currently the tiniest ascii cast is the one for the Schildbach "Bitcoin Wallet": https://walletscrutiny.com/android/de.schildbach.wallet/
Discussion
How do ASCII replays proof even less? Less than nothing can't make things worse.
I would not mind adding these if it wasn't so heavy on the git repo as it helps debug issues and is also a bit a demonstration of what's going on but I would never call them a proof.
In the end, reproducibility is something, people can attest to in order to extend trust or you have to reproduce it yourself which means you could as well just compile it from source always.
As we care about extending trust, we care about attestation and see jumping pixels more as a gimmick than anything else but if people like it, it's easy enough to record a screen session.