On Mon, Dec 16, 2019 at 04:48:41PM +0100, Florian Westphal wrote: > Pablo Neira Ayuso <pablo@xxxxxxxxxxxxx> wrote: > > > If its not a problem to display a non-restoreable ruleset > > > (e.g. unspecific 'type integer' shown as set keys) in that case > > > then the interger,width part can be omitted indeed. > > > > > > Let me know. For concatenations, we will be unable to show > > > a proper ruleset without the udata info anyway (concatentations > > > do not work at the moment for non-specific types anyway though). > > > > Indeed, what scenario are you considering that set udata might be > > missing? > > Any non-nft client/direct netlink user. Ah I see, as direct uapi users. > > We could still print it in such a case, even if we cannot parse it if > > you are willing to deal with. Just to provide some information to the > > user. > > If udata is missing, we only have the type available. > > If its a type with unspecific length (string, integer) we can use > the key length to get the bit size. > > But for concatenation case, it might be ambigiuos. > > So, I would remove the "type integer, length" format again so in > such case we would print > > type string > or > type integer. > > Users won't see this non-restoreable ruleset listed as long as the udata > is there. That's good enough by now, I think, thanks. Once we get more of those users, if they want some sort of interaction with nft, then we probably revisit this. Those clients might not even use nft for listing the ruleset at all.
