Went Through it With Brad
Johannes Ernst
jernst+openid.net at netmesh.us
Fri Nov 10 18:45:41 UTC 2006
I very much believe in backwards compatibility. I believe I'm saying
the same thing that Brad and David were saying in the original
message on this thread. This is a classical case of a tradeoff -- and
non-computer-interpreted version strings is the side I come down on
because I have tried the other and it can get very ugly when machines
mis-guess what happened.
On Nov 10, 2006, at 10:27, Pete Rowley wrote:
> Johannes Ernst wrote:
>>> 1) 9.1 in the openid.ns parameter talks about using this value in
>>> regards to compare with a lower version number of the protocol.
>>> Looking
>>> at things such as Jabber, it is nearly impossible to deal with
>>> version
>>> comparisons in any sane way. I'm not sure what the best way is to
>>> resolve this.
>>
>> In my experience, one should use a human-comprehensible versioning
>> scheme, but software should only do equals/does not equal
>> comparisons; no automatic "guesses" whether or not version 4 of
>> something is the same as version 2, or to what degree.
> That works if you don't believe in backwards compatibility :)
>
> Without the ability to determine the version in use is greater than
> some minimum version or the major version # is different to the one
> expected life gets pretty harsh.
>
>
> --
> Pete
>
More information about the specs
mailing list