I also got the impression that "spec" was also being used in the same
sense as SA, i.e., protocol/algorithm + configuration parameters.

Regardless, though, the point seems clear - reuse the style of terms
from other security docs ...

Joe

On 7/5/2016 12:18 PM, Black, David wrote:
> We're in semi-violent agreement on protocol vs. protocol specification.
>
>> I would suggest using different terms as needed in different places, and
>> copying the terms from IPsec where relevant here.
> The analogous IPsec term would be "protocol" for things like tcpcrypt and
> the TLS profile, with "protocol specification" as a natural term for a 
> document.
> As the draft currently stands, the result would be "encryption protocol" and
> "encryption protocol specification" both of which could be prefixed by TCP
> for precision.
>
> If needed, the IKEv2 term "transform" could be used to refer to elements
> of a TCP encryption protocol.
>
> Thanks, --David
>
>> -----Original Message-----
>> From: Joe Touch [mailto:[email protected]]
>> Sent: Tuesday, July 05, 2016 2:21 PM
>> To: Black, David; tcpinc
>> Subject: Re: [tcpinc] What to use instead of "spec"
>>
>>
>>
>> On 7/5/2016 10:52 AM, Black, David wrote:
>>> Like Kyle, I'm also uncomfortable with the use of the word "spec."
>>>
>>> How about "mechanism" instead?   That enables separation between
>>>     - Encryption mechanism (what is done to produce the bits "on the wire")
>>>     - Specification of an encryption mechanism  or encryption mechanism
>>>             spec[ification] (what's in the doc)
>>>
>>> The latter specifications are actually protocol specifications (e.g., they 
>>> include
>>> key exchange in addition to encryption).
>> FWIW,:
>>
>> 1) protocols are more than just what's "on the wire"
>>     They are the sum of the state machine on each end, the interfaces to
>> layers above and below, and the messages (on the wire).
>>
>> 2) a specification is a *description* of a protocol, algorithm, or mechanism
>>     It is distinct from the concept thereof and the implementations thereof.
>>     We tend to refer directly to the protocol/algorithm/mechanism
>> (unless we're updating a spec thereof).
>>
>> 3) I agree the doc uses "spec" to mean too many things:
>>     - TCP option variant
>>     - security transform (from IPsec, i.e., algorithm)
>>     - security association (from IPsec, i.e., transform and its
>> configuration parameters for a given TCP connection)
>>
>> I would suggest using different terms as needed in different places, and
>> copying the terms from IPsec where relevant here.
>>
>> Joe
>>
>>

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

Reply via email to