Re: [vwrap] Removing first name / last name assumptions?

[Barry Leiba <[e-mail]>]

On Mon, Apr 5, 2010 at 4:07 PM, Meadhbh Hamrick <[e-mail]> wrote:
> no. we're proposing taking it off the table because there is a very
> high likelihood that are valid first and/or last names in a particular
> implementation that include spaces and periods. this is the reason we
> came up with having first_name and last_name as separate strings in
> the transfer syntax, and not a single string identifying the avatar's
> name.
>
> why do we need to REMOVE support for Second Life from this protocol?

Well, here's the thing:

First: I've spent a lot of years designing protocols, and I think it's
a terrible idea to define a specific artifact of a specific
implementation in a new protocol.  We've done it many times before,
and it's pretty much always caused problems.  Sometimes those problems
eventually worked themselves out, but I'm quite sure it was a mistake
to go there.

Second: I don't see it as a question of "removing support for Second
Life", but one of looking at alternatives that are protocol-clean and
allow everyone to work with it.  And, by the way, I don't consider two
separate identifier fields, from which you're meant to pick which you
want to use, to be "protocol-clean".

Why not define the agent identifier in a self-parsing way?  Something
like this should work (written in ABNF):

   agent-identifier := delim 1*(id-part delim)
   id-part = string
   delim = 1char

...where the idea is that the delimiter is a single character that
appears as the first character (so it's self-identifying), and then
between ID parts.  The implementation would pick a character (which
could be a non-printable code point, such as 0x00 or 0x01, if
necessary) that doesn't appear in any identifier part.  Here's a
sampling of agent-identifiers, each within quotes just to make it
clear where they start and end:
   " Joshua Linden "
   ".Joshua.Linden."  (another way to represent the same thing)
   "/Joshua/Linden/"  (a third)
   "XJoshuaXLindenX"  (a fourth)
   " Melanie "  (an identifier with just one part)
   "rMelanier"  (the same, in another representation)
   ".Melanie."  (or less weirdly)
   " Barak Hussein Obama "   (here's one with three parts)
   "+Peter+Blair+Dennis+Bernard+Noone+" (a five-part one; extra credit
if you're old enough to know who he is)
   "~le roy~barr~"  (Meadhbh's example with an embedded blank)

This scheme only adds two characters to the identifier (the leading
and trailing delimiter, and we *could* eliminate the trailing one, if
people prefer, by defining agent-identifier as "1*(delim id-part)"...
I like having it), and seems to solve everyone's problem.

Barry, as participant