[DNSOP] Re: IPN and CLA RRTYPEs to support Bundle Protocol RFC9171

[[e-mail]]

I've lost lock on the ipn-scheme RFC, but my own assessment is that always sending a single 64-bit unsigned integer would be fine.  The application receiving the resource can figure out whether or not it wants to condense the value by representing it as two 32-bit integers in ASCII with leading zeroes suppressed and a period between the two.  Internally it's always going to be a 64-bitunsigned integer, from which a 32-bit "allocator" number can be obtained by simply shifting 32 bits to the right; if the result is zero then we're looking at an old-style IPN node number.

Scott

-----Original Message-----
From: Scott Johnson <[e-mail]> 
Sent: Monday, June 24, 2024 8:26 PM
To: Mark Andrews <[e-mail]>; [e-mail]
Cc: dnsop <[e-mail]>
Subject: Re: [DNSOP] IPN and CLA RRTYPEs to support Bundle Protocol RFC9171

Hi Mark,


On Tue, 25 Jun 2024, Mark Andrews wrote:

>
>
>> On 25 Jun 2024, at 10:32, Scott Johnson <[e-mail]> wrote:
>>
>> Hi Mark,
>>
>> On Tue, 25 Jun 2024, Mark Andrews wrote:
>>
>>> An obvious correction “LTP--v6” -> “LTP-v6”
>>
>> Aha!  Good eye.
>>
>>>
>>> For IPN why isn’t the wire format two network 64 bit integers?  That is 16 bytes.  Also 2^64-1 is 20 characters so 2 64-bit numbers separated by “." is 41 characters.  It’s not clear where then 21 comes from.
>>
>> EID is the basic unit of IPN naming, which is indeed two 64 bit integers separated by a ".". We are seeking to represent only the node-nbr component of an EID, as the service-nbr component is loosely analagous to a UDP or TCP port, for which there is one publicly defined service in the registry, and a collection of space agencies who lay claim to another chunk of them:
>> https://www.iana.org/assignments/bundle/bundle.xhtml#cbhe-service-num
>> bers As such, there is no gain in including the second 64-bit 
>> integer, representing service-nbr in the DNS records, and indeed, a loss of utility on the application level.
>>
>> The node-nbr component is presently, under RFC7116, a 64 bit unsigned integer.  There is a draft from the DTN WG currently making it's way through the IESG which will amend the IPN naming scheme. Perhaps I should add it to normative references?
>> https://datatracker.ietf.org/doc/draft-ietf-dtn-ipn-update/
>>
>> In effect it splits the node-nbr component into two-32 bit integers; Allocator Identifier and Node Number in the "Three-Element Scheme-Specific Encoding" of Section 6.1.2 over the above.  Section 6.1.1 describes the "Two-Element Scheme-Specific Encoding" method which retains the use of a single 64-bit integer.  Thus, a single 64 bit integer (20 characters) or two 32-bit integers (10 characters each) delimited by a "."
>> makes 21 characters maximum.  This preserves forwards compatibility with the proposed amended scheme, and does no harm if the scheme fails to achieve standardization.
>
> Or just 8 bytes on the wire with both possible input formats described. 
> Machines using the records will just be converting ASCII values to a 
> 64 bit integer.  We may as well transmit it as that.  Input validation 
> will need to do the conversion anyway to ensure both fields will fit 
> into 32 bits in the “.” separated case and 64 bits in the single value case.
> Length along is not sufficient to prevent undetected overflows.  The 
> only thing you need to determine is which format is the initial 
> canonical presentation format.  That can be changed with a later 
> update if needed.

I am tagging in Scott Burleigh, co-author of RFC9171 on this point for clarification. 
Section 4.2.5.1.2 of same indicates:

"Encoding considerations:
For transmission as a BP endpoint ID, the scheme-specific part of a URI of the ipn scheme SHALL be represented as a CBOR array comprising two items. The first item of this array SHALL be the EID's node number (a number that identifies the node) represented as a CBOR unsigned integer. 
The second item of this array SHALL be the EID's service number (a number that identifies some application service) represented as a CBOR unsigned integer. For all other purposes, URIs of the ipn scheme are encoded exclusively in US-ASCII characters."

Having already established that we are transmitting the node-nbr component only, and not a full EID, I am not sure we are restricted to using only US-ASCII.  ScottB, your opinion?  CBOR might also be an option, but that would place a higher burden upon implementers, I think.  Integer notation for wire format is fine by me.

>
>>> Limit CLA characters to Letter Digit Hyphen rather than the full ASCII range.
>>
>> It is possible for a node to support multiple CLAs on the same IP 
>> address and node number.  Will this change allow multiple, comma 
>> delimited values to be expressed in the CLA record?  If so, can you 
>> point me to an example so I can get the verbiage of the draft right?
>> If not, what do you recommend (in addition to my defining that in the 
>> draft)?  I like the idea of limiting the usable characters.
>
> Personally I would just use a TXT record wire format with the 
> additional constraint that the values are restricted to Letter, Digits 
> and interior Hyphens.  The input format matches the TXT record with 
> the above character value constraints.  The canonical presentation 
> form is space separated, unquoted, unescaped ASCII. This allow for 
> long records to be split over multiple lines.  Descriptive comments in the zone file.
> This take one extra octet over using comma separated values.

Sold to the man from ISC :)  This part works great; thank you!  Updated draft pushed to datatracker at https://datatracker.ietf.org/doc/draft-johnson-dns-ipn-cla/

Thanks,
Scott


>
> e.g.
>
> example inputs
>
> @ CLA ( TCP-V4 ; TCP over IPv4
>      TCP-V6 ) ; TCP over IPv6
>
> @ CLA “TCP-V4” TCP-V6
>
> Wire
>
> 06 ’T’ ‘C’ ‘P’ ‘-‘ ‘V’ ‘4’ 06 ’T’ ‘C’ ‘P’ ‘-‘ ‘V’ ‘6’
>
> Canonical presentation
>
> @ CLA TCP-V4 TCP-V6
>
>
>> Thanks,
>> Scott
>>
>>>
>>> Mark
>>>
>>>> On 25 Jun 2024, at 08:19, Scott Johnson <[e-mail]> wrote:
>>>>
>>>> Hi All,
>>>>
>>>> After reading the recent discussion about WALLET, I am hesitant to jump into the fray here, but this plainly is the correct group to help me get my logic and syntax right, so here goes:
>>>>
>>>> I submitted requests to IANA for IPN and CLA RRTYPEs, these representing the missing datasets necessary to make a BP overlay network connection from data found by DNS queries.
>>>>
>>>> For those not familiar, BP is a store and forward mechanism generally used in high latency situations where there does not exist constant end-to-end connectivity.  It was designed for deep space networking, however has network segments and application uses which overlay the terrestrial Internet.  There will arise similar use cases on the Moon (in the reasonably near future) and Mars whereby low latency, constant connectivity exists, thereby making use of DNS in these situations viable.
>>>>
>>>> My Expert Reviewer asked for an i-d, to clarify the requests, and that said i-d be sent to this list for review.
>>>>
>>>> Please find the approptiate draft here:
>>>> https://datatracker.ietf.org/doc/draft-johnson-dns-ipn-cla/
>>>>
>>>> Relevant IANA requests:
>>>> https://tools.iana.org/public-view/viewticket/1364843
>>>> https://tools.iana.org/public-view/viewticket/1364844
>>>>
>>>> I have the BP community also reviewing this, but they are generally in agreement as to use.
>>>>
>>>> Thanks,
>>>> Scott M. Johnson
>>>> Spacely Packets, LLC
>>>>
>>>> _______________________________________________
>>>> DNSOP mailing list -- [e-mail] To unsubscribe send an email 
>>>> to [e-mail]
>>>
>>> --
>>> Mark Andrews, ISC
>>> 1 Seymour St., Dundas Valley, NSW 2117, Australia
>>> PHONE: [phone]              INTERNET: [e-mail]
>>>
>>> _______________________________________________
>>> DNSOP mailing list -- [e-mail] To unsubscribe send an email to 
>>> [e-mail]
>
>
> --
> Mark Andrews, ISC
> 1 Seymour St., Dundas Valley, NSW 2117, Australia
> PHONE: [phone]              INTERNET: [e-mail]
>
>