Re: [5gangip] Reviewing GTP (was: re: Notes from today's meeting)

[Shunsuke Homma <[e-mail]>]

Hi Alex,

Thank you for your reply.

Regarding GTP packet format, if there are some differences between dump 
and the standard document or the standard document has several parts 
where is not clearly described, dump example would be helpful to 
understand actual GTP-based network structure. Especially, non network 
operator can hardly see such dump, and it would be useful. However, 
please note that it shouldn't be vender specific but should be general one.

In addition, I'd like to know architectural benefits and issues on GTP. 
For example, why current GTP-based network makes some anchor points or 
why ID/LOC protocols would be sperior to GTP-based forwarding mechanism. 
The clarification would be beneficial to improve ID/LOC and it would be 
also help to propose ID/LOC to mobile network.

Regards,

Shunsuke


On 2018/05/16 3:13, Alexandre Petrescu wrote:
> Thank you for the half-half direction.  I agree with the second half: 
> identify GTP issues.
> 
> I would like to express that this represents some work to do.
> 
> The easiest way to perform this work is to analyze a packet dump, and 
> describe it in English.  A packet dump is like a radiography of a 
> protocol: we see many things inside and understand what is going on, 
> without much speculation.
> 
> For this reason, I would like to ask you, if you ever have some chance 
> to come see a packet dump of GTPoUDPoIPv6, or even over IPv4, then that 
> would be of significant help.
> 
> For my part, I can obtain packet dumps of GTPoUDPoIPv4 captured on the PGW.
> 
> This will help us to make clear between which entities GTP runs (PGW is 
> sure, but UE?  SGW?  eNodeB?), its MTU and other aspects.
> 
> Even the simplest fact that somebody else than me shows a GTP packet 
> dump is very helpful - it shows there is interoperability.
> 
> That would be a significant help.
> 
> Alex
> 
> Le 15/05/2018 à 12:28, Shunsuke Homma a écrit :
>> Hi Alex,
>>
>> I think that there is no sense if its goal is republication of 
>> existing document (even if it's one of other SDO), but it would be 
>> useful if the document discribes clarification of issues/problems 
>> which GTP has in current mobile networks. And, it would be work which 
>> should be proceeded in 5gangip.
>>
>> Regards,
>>
>> Shunsuke
>>
>>
>> On 2018/05/14 19:44, Alexandre Petrescu wrote:
>>> Hi Mikael,
>>>
>>> Thank you for having read the draft.
>>>
>>> Le 14/05/2018 à 12:09, Mikael Abrahamsson a écrit :
>>>> On Mon, 14 May 2018, Alexandre Petrescu wrote:
>>>>
>>>>> Hi,
>>>>>
>>>>> I posted a draft about GTP.
>>>>>
>>>>> In order to access the draft one needs IPv6:
>>>>>
>>>>> http://petrescu.sandelman.ca/
>>>>
>>>> If you're going to make an IETF submission, please post it using the 
>>>> normal IETF ID way.
>>>>
>>>> If you're not making an IETF submission (which you don't seem to be 
>>>> since you've uploaded IETF draft formatted text to somewhere else), 
>>>> then please don't bring it up at IETF discussion groups and call it 
>>>> a "draft".
>>>
>>> Whether or not it becomes an IETF submission depends on the readers. 
>>> If enough people show interest then so it will be, otherwise no.
>>>
>>> At this time, I keep it on IPv6 access.
>>>
>>> As for not calling it a 'draft' - it is a good idea.  I will think 
>>> about it.
>>>
>>>> Also, after reading the document, it's impossible to understand what 
>>>> you're even trying to do with it, much less what you're trying to 
>>>> standardize (considering it says "standards track".
>>>
>>> Well, please see below the goals I have with this draft.
>>>
>>> Do you think GTP does not deserve be documented in an Internet Draft?
>>>
>>> The goals:
>>>
>>> GTP is a protocol that is not documented at IETF.  It is described in 
>>> 3GPP documents.  This lack of IETF documentation generates a lot of 
>>> problems.
>>>
>>> One such problem immediately visible is the following: people claim 
>>> IPv6-only access when actually their GTP is on IPv4 and just 
>>> transports IPv6.
>>>
>>> If there were an IETF document that said that GTP on IPv6 acts this 
>>> and that way then there would be no confusion.
>>>
>>> For example, we never confuse DHCPv4 for DHCPv6 or vice-versa, nor 
>>> RIPng for RIP, nor ND for ARP,  because we have distinct IETF 
>>> documents for each.
>>>
>>> Another aspect that may be interesting is the following: IANA 
>>> reserves a port number for GTP, but GTP is not specified at IETF.  
>>> This would probably be clarified, _if_ there were a document.
>>>
>>> As for the 'Standards Track' intention - I think it would be the 
>>> right track because GTP is in widespread use.  It is true that 
>>> GTP-IPv6 is not deployed anywhere, and that would not make it even a 
>>> 'Proposed Standard' status.  But in order to understand that, then 
>>> one may need to understand what is GTP-IPv6 first.
>>>
>>> Alex
>>>
>>> _______________________________________________
>>> 5gangip mailing list
>>> [e-mail]
>>> https://www.ietf.org/mailman/listinfo/5gangip
>>
>>
> 
> 


-- 
----------------------------------
Shunsuke Homma
<[e-mail]>
TEL: [phone]
FAX: [phone]

NTT Network Service Systems Labs.
Musashino city, Tokyo, Japan
----------------------------------