Incorrect parsing of high-precision timestamps about go-ssb

[staltz]

See the npm package monotonic-timestamp that Dominic created for this problem. It's used in ssb-validate

[arj03]

Yes, I that check seems wrong. In JS sometimes if you ask for time you'll get these fractional float values, that has always annoyed me. But they should all still be in the same ms format, and not in ns.

[boreq]

Exchanged some emails with @cryptix and this course of action was suggested:

• move the type to ssb-refs ssbc/go-ssb-refs#6
• tag a new go-ssb-refs version
• update https://github.com/ssbc/go-metafeed ssbc/go-metafeed#30
• update https://github.com/ssbc/go-gabbygrove ssbc/go-gabbygrove#3
• update https://github.com/ssbc/go-ssb #325

Opinions?

[boreq]

See the npm package monotonic-timestamp that Dominic created for this problem. It's used in ssb-validate

I am not entirely sure that this is the same problem. I am specifically referring to the fact that sometimes the value of the timestamp field is decoded differently. I understand that monotonic-timestamp just adds some extra stuff after the dot so the value itself is still in milliseconds.

That being said I don't understand why there is a requirement for the timestamp to always increase somewhere in the JS SSB codebase. If we measure in ms then obviously quick calls return the same results and there is nothing weird about this.

Also just to be clear "monotonic" means "non-decreasing" or "non-increasing" so getting the same value many times still means something is monotonic so again there is some confusion in the naming here (in the context of clocks in computer programs at least I believe).

[cryptix]

just while we are here anyhow: IIRC the increasing timestamps were intended to be used as a social "happened before" "proofs" but Dominic later realised that he could do that without taking the timestamp into account.

[decentral1se]

Nice! @boreq ssbc/go-ssb-refs#6 is merged and is out https://github.com/ssbc/go-ssb-refs/releases/tag/v0.5.2. Wanna do the update labours on #325 ssbc/go-gabbygrove#3 and ssbc/go-metafeed#30? Can also take a run if you like.

[decentral1se]

Oh wait, there is more: https://github.com/search?q=org%3Assbc+encodedTime&type=code?

Oh wait, that's old. I can't find anything on third look, so yeh, gonna close this off again 🙃

[boreq]

Perhaps @staltz knows more if the JavaScript stack also contains some kind of a fix for this quirk. If not I think we should drop this code.

[staltz]

Yes, I agree with you that the adherence to the definition of monotonic wasn't correct. Perhaps @arj03 can shed some light on why equal timestamps are a problem in SSB JS.

[boreq]

👍 But I think the problem here is not specifically with this monotonic timestamp library but rather the fact that sometime in the past there was a client which didn't encode the time as milliseconds but rather for example nanoseconds or something like that and this code that I found tries to correct for that by guessing that if a timestamp contains a dot in it then it means that it is in ns and not ms. This seems wrong as the code you linked produces fractional timestamps in ms. Therefore we can't use the dot to guess if it was in ms or something else.

[decentral1se]

If I'm reading this right, then you're GTG @boreq with your proposal?

I'm down to review/merge/migrate/click buttons.

[boreq]

Upstream doesn't indicate how to contribute patches but I will try sending a patch to the email address that I got from the git logs.

[cryptix]

welcome having this of my turf! i already 👍 the root PR and the rest is just renaming references so take this nod as going forward with the others once that is merged and tagged!

[boreq]

Ok all three remaining pull requests have been updated.

• https://github.com/ssbc/go-metafeed ssbc/go-metafeed#30
• https://github.com/ssbc/go-gabbygrove ssbc/go-gabbygrove#3
• https://github.com/ssbc/go-ssb #325

[decentral1se]

Nice, merged. Annnd #326 is also in. All good to go here? Nice coordination folks!

[boreq]

I think we can close this unless we find more references to encodedTime in some packages? I am on my phone but I can check again in the evening.

[decentral1se]

Ah, I think ssbc/go-ssb-room#372 needs to be done but we can follow up there.

I didn't see another reference in the go-ssb stack but feel free to re-open if you find one and we get to it!