OpenID status update

Martin Atkins mart at degeneration.co.uk
Fri Jun 3 04:18:21 PDT 2005


Kurt Raschke wrote:
> On Jun 3, 2005, at 6:02 AM, Ken Horn wrote:
> 
>> Is a middle ground to declare a version / protocol name with the
>> current impl?
> 
> 
> I think we need a version field anyway; obviously the protocol will
> change over time.  Major crypto changes notwithstanding, we need a way
> to make changes to the protocol without breaking everything.  Look at
> TypeKey, for example.  There is a version field in the TypeKey protocol,
> precisely so that servers know what version of the protocol clients can
> handle and can respond appropriately.
> 

However, TypeKey's version field wasn't added *until* there were
multiple versions. The lack of a version is taken to mean version 1.

Seems reasonable to me. If we wait until there *are* multiple versions,
we may instead have the opportunity to make the new versions gracefully
degrade, or signal the change by some other means which would avoid an
explicit version number. No need to get ahead of ourselves.

The current implementation is considered "unstable", so we can still
break backwards compatibility as much as we like until we say that it is
"finished". I think putting the consumer support live on LiveJournal
would be a good point where it would be considered "finished" because it
would be out in the wild, in the public eye.

In the mean time, I certainly don't have any problem with changes
breaking my little guestbook application! :)



More information about the yadis mailing list