Re: HTTP sasl draft in last call

On May 7, 2004, at 8:06 AM, Alexey Melnikov wrote:

> Nystrom, Magnus wrote:
>
>> Lisa,
>>
>> Thanks again for your review.
>>
>> The reason for introduction of 235/236 status codes is that they 
>> allow for
>> a clean introduction of a SASL security layer. With 200 alone, one 
>> must
>> ask when the SASL security layer actually goes into effect. Clearly 
>> not
>> all headers can be included in it, since for example there could be a
>> final WWW-Authenticate header from the server which is needed for the
>> client to calculate session keys etc. OTOH, any body clearly must be
>> protected by them. So, after some discussion, and while realizing 
>> that the
>> 235/236 will require an extra roundtrip, we felt the advantages 
>> outweighed
>> the disadvantages.
>>
>> Of course, I would be grateful for any suggestions on how to solve 
>> this
>> problem without the introduction of new status codes.
>>
> I agree with everything that Magnus said. I am wondering if this 
> explanation should be part of the document itself?

Unfortunately, I think the architecture that "necessitates" adding 
235/236 requires a rethink.  It's very un-HTTP and ill-defined to 
create a security layer in this way.  We could try to have some 
discussions about how to solve this differently - Roy suggested one 
approach, and there are obvious variations.

Lisa

Received on Monday, 10 May 2004 12:42:46 UTC