On Wed, May 13, 2026 at 02:39:26PM +0000, Job Snijders wrote:
> I think it is helpful to only consider CA material that can fit in
> USTAR archives and reject exogenous names. If 99 characters isn't

exogenous?

> enough, perhaps the CA should reconsider their filenaming scheme
> approach. I regret not having thought of a SIZE(5..99) contraint when
> the Manifest-bis RFC still was cooking as draft.

I'm not super keen on imposing a limit that hurts existing CAs even if
they ar super young.

The CA seems to be using krill (or a stack inspired by krill). That
krill chose to use this hex-encoded VRP notation nonsense for file names
is hardly the CA's fault. Even if the CA uses their own stack and chose
to mimick krill's behavior in that regard, I find it hard to fault them
for it - as much as I dislike these file names. That their INRs are what
they are probably isn't the CA's fault either.

Since it is well above the maximum krill ROA name (*), I can live with
the arbitrarily chosen limit of 255 (_XOPEN_NAME_MAX?) you suggested
later, although I'm not entirely sure what it buys us.

(*) As far as I can see, the longest ROA file name length is 130 in this
scheme: the longest possible 'notation' is 63 characters:

ffff:ffff:ffff:ffff:ffff:ffff:ffff:ffff/128 - 128 => 4294967295

8*4 hex digits, 7 separating colons, 4 for prefix, 6 for ' - max-length',
4 for " => " and 10 for AS number). Hex encoding doubles this to 126
and sticking ".roa" to the end lands us at 130:

666666663a666666663a666666663a666666663a666666663a666666663a666666663a666666662f313238202d20313238203d3e2034323934393637323935.roa

wonderful.