Hello Martin,

> I think the main question here would be deployability. The server is under 
> control of the person who decides what to allow, and clients get eventually 
> updated. But introducing restrictions like this might mean upgrading *users*.

Yes. However, at the same time, the character set which is currently
officially allowed on HTTP spec is ISO-8859-1 only, the ad-hoc handling of
non-ISO-8859-1 characters on HTTP authentication at this moment
is anyway not guaranteed to be interoperable.
(And because of ad-hoc handling of non-ISO-8859-1 characters,
 The latter half of ISO-8859-1 is also not well interoperable in reality.)

The HTTP community is going to adopt use of UTF-8 (with explicit
flag for using that), and it will anyway break backward compatibility
with non-ASCII usernames/passwords. (It will explicitly change
encoding of non-ASCII, ISO-8859-1 characters.)

It means that, at this moment, we CAN define the proper way of
handling for non-ASCII UTF-8 strings (with possible smallest
incompatibility with existing ad-hoc implementations),
and it is the last possible moment to do that
without severely breaking backward compatibility with
spec-conformant implementations.

Of course, it should not break any compatibility with existing
ASCII printable username/passwords.

2014-10-26 12:36 GMT+09:00 "Martin J. Dürst" <[email protected]>:
> Hello Peter,
>
> On 2014/10/23 11:38, Peter Saint-Andre - &yet wrote:
>>
>> [ Old thread alert! ]
>>
>> On 3/25/14, 12:52 AM, Yutaka OIWA wrote:
>
>
>>> My answers to the Peter's questions:
>>>
>>>> "Y                     u            taka  O   i    w     a"
>>>
>>>
>>> Current HTTP allows it, and we don't need to reject that mostly.
>>
>>
>> Does the HTTPAUTH WG want to support everything that HTTP currently
>> supports, or is it open to restricting things a bit more?
>
>
> I think the main question here would be deployability. The server is under
> control of the person who decides what to allow, and clients get eventually
> updated. But introducing restrictions like this might mean upgrading
> *users*.
>
> Essentially, the server operator will have to mechanically check which user
> names won't work anymore in the new protocol, and then contact these users
> and ask them to change their user name or tell them that their user name was
> modified. Not all users might like that.
>
>
>> Has there been further discussion about this in the HTTPAUTH WG? I'm on
>> that mailing list and I haven't seen much discussion on this topic.
>
>
> Me neither.
>
>
> Regards,   Martin.

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

Reply via email to