Re: [Unbearable] Fwd: I-D Action: draft-ietf-tokbind-ttrp-01.txt

[Vladimir Dzhuvinov <[e-mail]>]

On 04/08/17 00:14, Brian Campbell wrote:
> I agree that standardizing the header names significantly increases the
> likelihood that implementations will properly scrub them.
>
> I'm told that this mod_token_binding module for Apache
> <https://github.com/zmartzone/mod_token_binding>, for example, was recently
> updated to do the header sanitation automatically in the course of
> processing rather than relying on configuration directives of other modules
> to overwrite the headers. So it can't be misconfigured, at least not with
> respect to this issue. Standardizing the header names allows and suggests
> for the sanitation to be an implementation thing rather than a
> configuration thing, which is a good thing.
If the proxy implementers get on board with this, as you mention, it
will be great. It helps that the implementations are relatively few.

I wonder, has there been anywhere work on standardizing a header pattern
that can only exist between a reverse proxy and a backend server?


> As John said, there were a number of suggestions in Prague and on-list
> around how to prevent/detect header injection in more of a fail closed way.
> Vladimir's specific idea didn't come up but there were conceptually similar
> ideas.
>
> I maintain that it would be very inappropriate for this draft to define a
> one-off mechanism for a backend server to detect client injection of the
> headers. The potential problem of client header injection is not at all
> unique to the functionality of the draft so the draft shouldn't define a
> unique solution.
+1
>  If there's the will and/or appetite to develop a common
> standardized mechanism detecting/preventing client header injection through
> reverse proxies, I'd be happy to help contribute to that effort.
>
>
> On Thu, Aug 3, 2017 at 9:31 AM, Bill Cox <[e-mail]> wrote:
>
>>
>> On Thu, Aug 3, 2017 at 6:32 AM, Vladimir Dzhuvinov <
>> [e-mail]> wrote:
>>
>>> The downside of having set header names is that attacks on badly
>>> configured TLS proxies become much easier to carry out. This is my concern
>>> about the TTRP spec in this form.
>>>
>>> Thanks,
>>>
>>> Vladimir
>>>
>> I think most coders adding the standard header will the RFC, and will be
>> more likely to remember to scrub the header.
>>
>> One question about the spec: Why must the "sec-token-binding" header be
>> removed?  I did that originally in an implementation, and was asked to stop
>> "molesting the headers".
>>
>> Bill
>>
>> _______________________________________________
>> Unbearable mailing list
>> [e-mail]
>> https://www.ietf.org/mailman/listinfo/unbearable
>>
>>