[db-wg] In-band notification mechanism?

 Hi Aris
This is what I was thinking you were looking for. I just wanted to be clear. Knowing how the database is structured I can think of ways of doing this, but it would be for the RIPE NCC to assess feasibility. I think it may be easier to do it as a service to members than making it a more general feature for anyone. 
Does anyone else agree on the need for such a feature? If so perhaps we can create a new Numbered Work Item.
cheersdenisco-chair DB-WG

    On Saturday, 23 March 2019, 19:21:01 CET, Aris Lambrianidis <aris.lambrianidis at ams-ix.net> wrote:  
 
 Hi Denis,

We, and other IXPs, create filters (prefix-lists) for services such as route servers, by parsing aut-num and as-set objects from IRR databases, such as the RIPE database, using tools such as bgpq3. 

Right now, to the best of my knowledge, the only way to maintain those filters up to date for all of our route server peers, is to periodically poll IRR databases for changes. 
IMO it would seem more efficient if the database itself notified us of any changes, rather than us constantly asking the same question(s).

Does this make sense?

That said, I can also think of other use cases in which interested parties having no direct relationship to certain objects and their maintainers are interested in finding out of any changes, 
especially in the field of research, but let me not delve into this and keep things simple for the time being.

Kind regards,
Aris

ripedenis at yahoo.co.uk wrote on 23/03/2019 02:26:

  Hi Aris
Can you clarify one point about this. Are you saying you want to be notified if someone changes their data that you have no direct relationship with? So if I maintain a set object and you are not part of my company and have no direct business relationship with me and I have no idea who you are, but if I modify this object you want to be notified?
cheersdenisco-chair DB-WG
     On Saturday, 23 March 2019, 01:02:48 CET, Aris Lambrianidis via db-wg <db-wg at ripe.net> wrote:  
 
 Hi Wilfried,

Thank you for the effort in helping out!

Unfortunately this will not do as:

1. It notifies via an "out-of-band" method (i.e. email). This makes
it difficult (but not impossible) to handle with automation. 
Nonetheless, the
more elegant way would be through an API leveraging a push mechanism.

but more importantly:

2. the "notify:" attribute has to actually be configured with an address 
of the
interested party for it to work.

However I'm looking for mechanism for interested parties to be notified of
any changes in objects independently to what the maintainer has configured
as a notify address.

Kind regards,
Aris


Wilfried Wöber wrote on 22/03/2019 21:50:
> Hi Aris!
>
> Is this what you are looking for?
>
> https://www.ripe.net/manage-ips-and-asns/db/support/documentation/ripe-database-documentation/notifications/9-2-notification-messages/9-2-1-notification-attributes
>
> I may be off-track, of course :-)
> Wilfried
>
> On 22/03/2019 20:29, Aris Lambrianidis via db-wg wrote:
>> Dear all,
>>
>> Back in the day, RFC1996 introduced the NOTIFY mechanism in DNS, which significantly helped with information propagation delay,
>> as it facilitated the transition from a pull (poll) to a push (interrupt) model.
>>
>> The problem we, as AMS-IX, are facing is quite similar when it comes to polling the RIPE database for changes. This seems
>> inefficient.
>>
>> Although the analogy breaks down quickly, as there are no RIPE database "clients" similar to DNS slave servers
>> parsing NOTIFY messages, we would love to see any RIPE API created or extended, or any other mechanism implemented by which
>> a client "registers interest" for any objects it wants to be notified of changes.
>>
>> As a simple example, if we were to "register interest" (e.g. via a REST POST or PUT method) for the AS-AMS-IX-SET as-set object, we would be
>> programmatically notified whenever the "last-modified" field of the as-set was changed.
>>
>> Based on the above, I have 3 questions:
>> 1. Does something like what is described above already exist?
>> 2. If it doesn't exist, would others be interested on such functionality?
>> 3. If it doesn't exist, while knowing that this is only a high level overview of the concept and many details are missing, is this generally feasible?
>>
>> Kind regards,
>> Aris Lambrianidis
>> AMS-IX
>>


  

  
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.ripe.net/ripe/mail/archives/db-wg/attachments/20190323/dc66ad4e/attachment.html>