rpki-client 9.6 backwards compatibility

[jurraca]

rpki-client 9.6 introduces a better way to parse ROAs which we would like to leverage (see #97), but would make kartograf not backwards compatible. While we package the newest rpki-client in our Nix developer environment and build, a user is still free to bring their own developer environment with Python dependencies and the rpki-client of their choosing, and we want to continue to enable that.

One solution would be to provide only fixes on a 0.4.x branch with rpki-client v9.6 and the current RPKI validation logic, and a 0.5.x branch with rpki-client v9.6 with the new -p thread logic on which active development happens.

This perhaps gives us the opportunity to consider removing the "experimental" warning with 0.5.x.

Summary:

kartograf 0.4.12 "al-idrisi", rpki-client 9.5

• batched parsing
• deduping on valid_since, valid_until

kartograf 0.4.13 (not released), rpki-client 9.6 (see #102)

• upgraded to rpki-client 9.6
• no logic changes from 0.4.12

kartograf 0.5.0, rpki-client 9.6

• parsing with -p
• not deduping on valid_since, valid_until

Opening discussion regarding plans around this and best way to proceed.

[fjahr]

The one aspect that contributes to the upgrade being problematic that isn't mentioned here, is that new rpki-client versions aren't usually rolled out to all package managers and OSs very quickly. Building the release yourself can be a bit challenging. This is why we can't just upgrade to a 0.5.0 today and require everyone to use rpki-client >=9.6. See https://repology.org/project/rpki-client/versions.

upgraded to rpki-client 9.6

I think saying the 0.4.x versions are "tested with 9.6" and don't break behavior. And 0.5.x is "leveraging breaking 9.6 features".

I think I might have had some different thoughts that I don't remember but right now it seems like we should just hold off of the 0.5.0 for a few more weeks and then we might be good to go. I would say we release 0.4.13 before the next run and then we'll be able to see in the logs if the participants already had access to rpki-client 9.6. We could use that issue also to ask if anyone would have issues with a forced upgrade any time soon.