Re: [OAUTH-WG] Mandatory-to-implement token type

[Barry Leiba <[e-mail]>]

>> So an MTI token type + no client preference is equivalent to there
>> only existing one token type.
>
> Maybe.
>
> However, no MTI token type + no client preference = no interop.

I think this exchange, coupled with what you Stephen said in Taipei
(noting that "mandatory to implement" doesn't mean "mandatory to
use"), points out a basic difference between how Stephen is thinking
about this and how at least some of the WG participants are.  Let me
see if calling this out helps:

Stephen is thinking that code will be reused.  Perhaps there'll be a
limited number of coded toolkits, and their code will be used to
implement various authorization servers, etc.  That's the way a lot of
Internet code is done today.  In that case, Stephen posits, if we tell
the toolkit writers that they MUST implement X, then at least when X
is being used, implementations based on the different toolkits can
always interoperate.  Lacking that directive, a toolkit that only
implements X will be unable to interoperate with a toolkit that only
implements Y.

Others are thinking that deployed code will be specific to a
particular environment, and it doesn't help at all for that code to
support X if the environment demands Y.  There's no interop benefit
from the "MUST implement X" directive if the code has been written
specifically for that environment, and, as Mike T says, it only
results in untested code and wasted development time.

I see both sides of this, but in the end I think what Mike says is
spot on, here.  I think it would be very nice to strongly encourage
toolkit writers to support all the available options, but there's
absolutely no reason to require, and no benefit in requiring,
purpose-built code to do anything of the sort.

Maybe what would work best is some text that suggests what I say
above: that toolkits intended for use in implementing OAuth services
in general... implement [X and/or Y], and that code written for a
specific environment implement what makes sense for that environment.
It seems to me that to require any particular implementation in the
latter case is arbitrary and counter-productive, and doesn't help
anything interoperate.  Whereas general-purpose toolkits that
implement everything DO help interop.

Barry, as participant