EventID incorrectly inherits from DomainSpecificString

[clokep]

Not super happy with the startswith but EventID.is_valid is broken for non-V1/2 event IDs so unsure how to proceed.

Originally posted by @Fizzadar in #14632 (comment)

Modern event IDs are now hashes and do not include a domain name, so they shouldn't inherit from DomainSepcificString.

[reivilibre]

good spot. We should define a new string type for this with a corrected grammar?

[clokep]

good spot. We should define a new string type for this with a corrected grammar?

Yep! I don't know if there's anything that would break if it no longer has the same interface as DomainSpecificString.

I think the grammar is just $[...characters...] unfortunately.

[reivilibre]

That wouldn't surprise me, I guess. Probably null bytes might be accidentally excluded because of database support but everything else might be acceptable.
v3+ event IDs have a more strict grammar owing to the fact they are an encoded hash, so we could define an alphabet for those. For the non-v3+ event IDs, we can enforce that there is a domain.

[DMRobertson]

v3+ event IDs have a more strict grammar owing to the fact they are an encoded hash, so we could define an alphabet for those.

I think they also have an expected length?

For the non-v3+ event IDs, we can enforce that there is a domain.

Similarly, I think the spec might define a maximum length of an event ID here