Re: [tsvwg] draft-ietf-udp-options issues from IETF 104

[Joe Touch <[e-mail]>]

On 2019-07-15 12:46, Tom Herbert wrote:

> On Mon, Jul 15, 2019 at 10:40 AM Joe Touch <[e-mail]> wrote: 
> 
>> ...
>> 
>> Sorry, but you're not following where were are with the UDP options.
>> 
>> 1) we already agreed (in the past week) to make the OCS field mandatory
>> 
>> 2) OCS always covered the entire option space
>> 
>> 3) we already agreed that there would NOT be any surplus space beyond the options, i.e., that options would take up the whole surplus space
>> (anyone who wants to use surplus space should coordinate with the UDP options using that mechanism, as I noted when evaluating your proposal as not necessary)
> 
> Please explain how LITE mode works robustly if the UDP checksum is set.

LITE isn't robust, by design, but... 

- we have another thread about that issue, but we're NOT DISCUSSING THAT
RIGHT NOW 

- IF WE WERE DISCUSSING THAT NOW, I'd note that OCS could do the
following: 
    1- SHOULD be used, but MAY be set to zero (at the user's own
discretion and potential peril) 
    2- LITE ***might*** use OCS set on transmission but ignored on
receipt (to *allow* errors, because LITE ISN'T ROBUST **by design**) 
    3- LITE+FRAG can do as LITE does (#2), with a reassembly checksum
instead (ESP. given the individual checksums null-out with the data,
that may be more appropriate) 

However, we're still not talking about that yet. 

Joe