On 2/13/20 8:31 PM, Dave Cridland wrote: > On Wed, 12 Feb 2020 at 16:54, Georg Lukas <ge...@op-co.de > <mailto:ge...@op-co.de>> wrote: > Persistence: certain PEP implementations from the past, which are still > distributed by major OS platforms, chose to implement PEP in a > non-persistent way, only keeping data in RAM. I know we are deep in > implementation-defined behavior territory here, but a warning to > developers might be appropriate. > > > No idea what to say here. Feels like "Clients are RECOMMENDED to use > servers which are not totally shit".
I wonder if the right conclusion to draw here, may be that we need a feature announcement if the PubSub/PEP service will persist nodes or not (and how many items?). In the past, some PubSub/PEP implementations where bootstraped without persistence support, likely because it is easier to get a pure in-memory implementation done first. And it is probably safe to assume this will also happen with some new implementations in the future. Mind that some PubSub/PEP use-cases are perfectly fine without item persistence. Hence why not add such a feature [1] and have XEPs like xep402 require clients to check for the existence of this feature? - Florian 1: Probably to xep60? _______________________________________________ Standards mailing list Info: https://mail.jabber.org/mailman/listinfo/standards Unsubscribe: standards-unsubscr...@xmpp.org _______________________________________________
