On 23/08/16 19:26, Jaromir Talir wrote:
  On Tue, 2016-08-23 at 16:23 +0200, Piotr Przybył
wrote:
  Hello cz.nic Team
 I have a few questions related to database migration of FRED if you
 don't mind.
 ====
 When trying to download the latest sources from 
https://fred.nic.cz/p
 age/2904/download/#source I've
 realised, that they're not available. E.g. the latest migration
 scripts are available in
 fred-db*deb, but the tarball is not a corresponding one.
 Could you please try to release latest sources in tarballs? 
 The last version I've uploaded on the website is 2.23. We have released
 small upgrades 2.24 and 2.25 but there is an issue with building them
 on Fedora 24 so I'v postponed next website update to 2.26 that will be
 available for recent Fedora system. I believe this will be available
 next week and I'll do website update immediately after that. 
What I had in mind precisely was that the page lists e.g.
https://fred.nic.cz/files/fred/sources/fred-db-2.21.6.tar.gz, which doesn't contain
notification_queue for instance. I discovered the SQL upgrade script installed by the
package, but
can't see it in sources tarball.
   ====
 Could you briefly describe what's the purpose of tables:
  contact_address
  contact_address_history
 As far I can tell after analysing the sources, these tables are
 populated by a functionality called
 from MojeID ( =your special NIC registrar to keep domain contacts
 defined at central level, not each
 registrar's level).
 
 There is a concept of single address in FRED. In our mojeID registrar
 we have a concept of multiple addresses (mail address, permanent
 address,...). We decided to move this concept down into FRED and
 implemented multiple addresses in FRED database. The next step was
 supposed to be propagate this multiple address manipulation into EPP
 but this was postponed for a while. I believe we will get to his soon. 
 
I believe this can make a fair usage in case of a registrar. E.g. you create an invoice
with address
A, print it and mail to address B. However, I'm not convinced that this makes sense in
case of a
registry which sends nothing at all and the only address is supposed to be "the legal
address". I
understand your situation and efforts with MojeId, I'm only wondering if other
registries will
utilise that. (Although 1 is still a valid case of having N choices.)
When it comes to changes addresses it might be nice for some to make the street not
obligatory
field, because there are countries in which the street address can be missing and it's
perfectly
fine. ;-)
  ====
 What are the following tables for? Or: How the checks of contacts are
 working? Is is something
 related to checking if a contact's address fields are valid?
  contact_check
  contact_check_history
  contact_check_message_map
  contact_check_object_state_request_map
  contact_check_poll_message_map
  contact_test_result
  contact_test_result_history
  contact_testsuite_map
  enum_contact_check_status
  enum_contact_check_status_localization
  enum_contact_test
  enum_contact_test_localization
  enum_contact_test_status
  enum_contact_test_status_localization
  enum_contact_testsuite
  enum_contact_testsuite_localization 
 That would be loooong explanation :) I believe this will be part of new
 documentation that Lena is working on. I talked about this briefly
 during ICANN Techday
  
https://fred.nic.cz/files/fred/FRED-Validation.pdf slide 12,13
 (selective contact validation).  
 
So my guess that this is for contact/address validation was right. ;-)
I'm not aware of the law you must obey/implement in your system.
This seems to be like a policy forcing every contact to be nice and valid for all domains,
even if
they're not causing any trouble (like having sites which might be illegal in any part)
and BEFORE
they cause trouble.
Are you forced to do that? Maybe you're not allowed to assume that each
registrant/admin is honest
and peaceful "until proven guilty", or even "accused"?
  ====
 There's a table notification_queue. I can't find any usage of it
 (maybe because not all tarballs are
 up to date). What is it for? 
 I was going to mention this in mail summarizing changes in FRED-2.24,
 2.25 and 2.26. There is a new concept of asynchronous notification in
 FRED-2.25. In previous versions when registrar issued EPP command that
 involved sending notification email to registrant, EPP command was
 waiting for notification subsystem to create this email. Asynchronous
 notification means that there is only small record in
 notification_queue table that notification should be send and
 asynchronous script will create this email in separate process. The
 advantage is greater speed of EPP and stability because when subsystem
 for notification email creation is not available EPP is not locked. 
 
Oh yes, this makes a perfect sense.
  Do you replicate this table using Slony? (Because
it has neither
 primary nor unique key.) 
 We are not using Slony anymore, we have changed to internal streaming
 replication about 4 years ago. Maybe that's why we missed to add
 primary key. I'll ask my colleagues to add this primary key.
  
 
Thank you for your answers.
Best regards
Piotr