On Fri, 25 Mar 2011 12:21:10 -0400, Mark Rejhon <[email protected]> wrote: > -- Wild idea out of the blue: what about the idea of a different > standard (third standard) called a message indexing standard, a method > of referencing a specific historical line, that both your standard and > possible future versions of my standard, can take advantage of?
Using the 'id' attribute on message stanzas would be a clean way of indexing messages... if only that attribute were guaranteed to always exist on such stanzas. Unfortunately, the 'id' attribute is only mandatory on iq stanzas. I can see two approaches for solving this: 1. Only messages with an 'id' attribute can be later referenced/edited. Smart clients would always insert that attribute on messages. I like this approach because it requires no change to any spec, and doesn't introduce compatibility issues. Only messages without an id couldn't be edited later. We need to decide whether this is a problem, and whose. For this feature (correcting one's own message), I would say that this is the emitting client's, so I don't consider it an interop issue. 2. Modify one of the specs somewhere in the chain to require id attributes on messages. What spec, and how? Does anybody see another approach? FWIW, let's not forget that there have been, and will be, other situations where an id attribute on messages would be of great help. For example (but there might be others), being able to quote/reference/point a previous message in a conversation. In some use cases, not including this attribute could be someone else's problem. For instance, a flag-this-message-as-spam feature would definitely not be happy with optional id attributes. Who would include an id on their spam so that it can be easily flagged? ;) Cheers. -- David
