Re: [vwrap] [ogpx] type-system : version tag and handling unknown tags

[Mark Lentczner <[e-mail]>]

In trying to think thought the scenarios where a version field might be workable, I'm coming up short:

A v.n only reader encountering v.n+1 could only make two choices: Ignore everything it doesn't understand and assume that the semantics of the content aren't affected, or reject the transmission entirely. We'd have to choose now how it should react.

If it is ignoring things, then it seems that any v.n+1 extension would have to be defined like "If you understand v.n+1, then ignore some (or all) of the v.n stuff and substitute this richer data." But at that point, we are mostly using this version stuff as a way of stuffing two or more versions of the data in the same container. Extending this beyond one or two revisions is likely to get out of hand.

If it is rejecting things, then this has little use here, since there must be some signaling and protocol for negotiation to handle this case in some outer scope (HTTP, say). But then, why not just embed the version information at the higher scope (MIME-type, say)?

The "ignore by preserve" semantic works well for processing streams, and in cases where the vast bulk of the data is understood, and the ignored, foreign markup, isn't semantically important. VWRAP's use of LLSD doesn't seem to fit that model.


Mark Lentczner
Sr. Systems Architect
Technology Integration
Linden Lab

[e-mail]

Zero Linden
[e-mail]