On 2015-01-08 14:37, Edwin Mons wrote: > On 08/01/15 14:18, Kurt Zeilenga wrote: >>> On Jan 7, 2015, at 9:27 AM, Ralph Meijer <[email protected]> wrote: >>> >>> On 2015-01-07 16:15, Adrien wrote: >>>> Hi, >>>> >>>> On 01/07/2015 03:42 PM, Edwin Mons wrote: >>>>> Hi all, >>>>> >>>>> XEP-0060 lists both delete-items and retract-items for the same feature, >>>>> <retract/>. delete-items was added in the last revision, but it looks >>>>> like an error to me. I think 7.2 should be revised, and one of the two >>>>> features (likely delete-items) should be removed. >>>> yes you're right. At least that's what I have been told ("delete-items >>>> is the surplus one") when I asked. >>> I forget why it was added in the last version of this XEP, but it surely >>> wasn't missing as mentioned in the changelog, as we had 'retract-items' >>> as a feature for a long time. Reading the text around it, I feel it is >>> confusing, too. I'd rather go with talking about 'retracting items' and >>> 'deleting nodes' thoughout, and not talk about 'deleting items'. >>> >>> I think that many (all?) clients ignore this feature for discovery >>> altogether, so what about making 'delete-items' a thing that servers >>> SHOULD also advertise along with 'retract-items', but have clients >>> depend on 'retract-items' exclusively? >> Personally, I rather not add cruft like this. Has any server ever >> advertised this? Has any client ever relied on this? And, if it’s only >> SHOULD, no client can rely on it… Better, IMO, to just require one >> <feature/> to be advertised per feature. >> > > Eh, I've seen delete-items advertised in the wild, because M-Link > advertises it.
Hi, Edwin also tells me he has been working on this, but M-Link doesn't currently advertise the (older and IMO proper) feature called retract-items. Idavoll only sends the latter, and Prosody sends both. I haven't done a thorough check on other implementations. As such, my suggestion was meant to be an attempt to have backwards compatibility, which I am a strong supporter of. That said, I've not found evidence of clients having their use of the retract protocol depend on either of these features being advertised. Finally, I'm not convinced that retraction should be RECOMMENDED, and think the current registration of retract-items being OPTIONAL is just fine. -- ralphm
