Re: [[e-mail]: New Version Notification for draft-jones-6man-historic-rfc2675-00.txt]

[Brian E Carpenter <[e-mail]>]

On 09-May-19 05:26, Bob Hinden wrote:
> Hi,
> 
>> On May 8, 2019, at 9:07 AM, Ole Troan <[e-mail]> wrote:
>>
>>> Hello 6man,
>>>
>>> We have put this together to change the status of RFC2675 to Historic
>>> and would like to request discussion in the working group.
>>
>> IPv6 jumbograms was intended for some super computer inter connect with a massive MTU.
>> I don't know of any use of it, but is it harmful if the specification is left there in place?
>>
> 
> Are there any current network interfaces that can support packets larger than 65,535 octets?   As I remember, it was intended for network interfaces like hyperchannel that could support packets larger than 65,535.    Did we every define an IPv6 over Hyperchannel specification?
> 
> As a co-author of RFC2675, I am OK with making it historic.   It was very limited in use when first defined, and hasn’t gotten better. 

I disagree. This was designed for a specific usage scenario that was of some concern to me at the time: use within data-intensive data centres handling distributed applications dealing with very large datasets. (At CERN we were already thinking about the future LHC datasets back in the 1990s.) There was hardware in use at the time (HIPPI) that allowed exceedingly large frames. Now things haven't worked out exactly as we expected, but scientific datasets are only getting bigger as time goes on. Jumbograms might be needed in future data scientific data centres.

I don't think we should obsolete this possibility. It is very much a case in point of a limited domain protocol that will never be used over the open Internet. If unused, it's harmless. General-purpose transport protocols don't need to consider it.

It's absolutely reasonable to make it an OPTIONAL feature of IPv6 implementations, which is already the case in practice. An RFC that clarifies it as optional would be a fine thing.

Oh look, we have an RFC that says it's optional:

"  Jumbograms are relevant only to IPv6 nodes that may be attached to
   links with a link MTU greater than 65,575 octets, and need not be
   implemented or understood by IPv6 nodes that do not support
   attachment to links with such large MTUs."

That's RFC2675. We're fine as we are.

If the Transport area wants to make a policy that transport protocols do not need to support jumbograms, that's fine too.

Regards
     Brian