On parentheses and changes

I don’t have much time here, but I wanted to make a quick note about parentheses and cross references.

First, I’m not able to keep up with all the discussions – so Drummond, while I sounded like I agreed with the proposal you were making a week or two ago, I think things have moved so fast that I don’t feel like I even understand what it is I agreed to at this point.

Second, I’m very concerned that we’re treading into areas (changing xref syntax) that we might find ourselves deciding later on we could have gone either way on. Yes, right now, the particular concerns about human-friendliness have taken over the conversation, but we also have machine friendliness to consider. Steve’s advocacy of the computer’s interest (hah!) in the form of “abstract” syntax is addressing this area – though I have to admit I haven’t been able to follow in great detail. Personally, I never thought humans would be constructing cross references, so I’m finding these “easier for humans” arguments a little unpersuasive this late in the game.

Thirdly, we have to be careful about changing things that people are already using. I’m worried on a couple of angles. First, we have deployed code out there, and while nobody has spoken up about using the current syntax, I’m worried there are people out there who may be, and aren’t aware of the scope of the changes being proposed. Second, there’s an XRI that’s pretty widely deployed *right now* that contains an xref that I believe would become an “oddity” by the new syntax architecture – namely “xri://$xrd*($v*2.0)” Every OpenID identifier has an XRDS document using this XRI as a namespace identifier. Third, we are already seeing folks prepare derivative documents based on our documents because the folks believe our documents are too complicated – that’s fine, but if we start going changing our stuff, I worry that a) we’ll lose control of the XRI specs and b) we’ll lose credibility altogether for XRI if we change stuff after stuff has been deployed out in the wild.

My last point needs to be re-emphasized. I think we’re reaching a point where enough people are relying on our stuff that we can no longer change it willy-nilly and be able to rely assertions from TC members that nobody is using a particular feature or construct. We claim that nothing we’re doing is “backwards incompatible” but the more basic stuff we muck with, the less I’m confident that we really understand how our changes might ripple through implementations and adoption assumptions made based on reasonable reading of earlier drafts.

-Gabe

xri message