I added:

  Parameters sent without a value MUST be treated as if they were omitted from 
the request.

This should cover it.

EHL


From: Andrew Arnott [mailto:[email protected]]
Sent: Tuesday, June 15, 2010 9:49 AM
To: Eran Hammer-Lahav
Cc: OAuth WG ([email protected])
Subject: Re: [OAUTH-WG] Clarification on whether arguments can contain empty 
values

Thanks, Eran.  I can live with that.

Do you think it would be a good idea to state in the spec that parameter values 
may be empty?  For instance for this one:
expires_in
OPTIONAL. The duration in seconds of the access token lifetime if an access 
token is included.
Maybe I'm taking it to literally, but I read this to include "if the parameter 
is present, its value WILL be an integer".  But if somewhere in the spec it 
said "all optional parameters may be present but empty", then that defeats by 
incorrect interpretation.

Again, I can live with an email from you, but it might help other implementors 
who come along next year to get it right the first time if they're reminded in 
the spec that parameter values may be empty.  At least it would have helped me 
a few years ago in the OpenID spec. :)
--
Andrew Arnott
"I [may] not agree with what you have to say, but I'll defend to the death your 
right to say it." - S. G. Tallentyre

On Tue, Jun 15, 2010 at 9:43 AM, Eran Hammer-Lahav 
<[email protected]<mailto:[email protected]>> wrote:
The best way to address this is to write more resilient servers. Servers should 
accept empty optional parameters.

EHL

From: [email protected]<mailto:[email protected]> 
[mailto:[email protected]<mailto:[email protected]>] On Behalf Of 
Andrew Arnott
Sent: Tuesday, June 15, 2010 6:56 AM
To: OAuth WG ([email protected]<mailto:[email protected]>)
Subject: [OAUTH-WG] Clarification on whether arguments can contain empty values

Can we get some clarification into the spec as to whether optional parameters 
can be present but empty?  Particularly parameters such as tokens that 
obviously cannot be meaningful when having an empty value.  This was a muddy 
issue in the OpenID spec, where some implementations would include empty 
parameters rather than just omitting them, breaking other implementations that 
would expect that if the parameter is present it ought to have a meaningful 
value.

My own vote: parameters must have valid values (non-empty) if they are present, 
unless they are opaque strings (like client state) that the remote party 
doesn't have to do anything but imitate back anyway.

--
Andrew Arnott
"I [may] not agree with what you have to say, but I'll defend to the death your 
right to say it." - S. G. Tallentyre

_______________________________________________
OAuth mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/oauth

Reply via email to