Re: GNU MediaGoblin

We plan to use URLs as IDs for 
media/comments/collections/activities/etc. and acct:username@host.tld 
for users e.g:

     acct:tsyesika@mediagoblin.mydomain.co.uk
     http://mediagoblin.mydomain.co.uk/api/comment/1

The internal IDs should remain completely internal and the API will only 
expose the global URL IDs. If we were writing MediaGoblin from scratch 
today we probably wouldn't have integer based internal IDs.

We currently have some basic support in our models for serialization to 
JSON you can see this: 
https://gitorious.org/mediagoblin/mediagoblin/source/1c8f52daee64a337525ad54a549c6a0e60bb11f0:mediagoblin/db/models.py#L532-594

Some info on what the data those attributes on the User model hold:

     id = The internal integer based ID (legacy)
     public_id = federated URL like ID which will be used when 
interacting with other servers
     slug = slug based on the title which is used to create the URL that 
the user will see (i.e. not the API link but the user facing website)

The "url" parameter I'm presuming is on the User? This is the users 
personal site that will be displayed on their profile.

Thanks,
Jessica Tallon.

------ Original Message ------
From: "☮ elf Pavlik ☮" <perpetual-tripper@wwelves.org>
To: "Jessica Tallon" <jessica@megworld.co.uk>; public-socialweb@w3.org
Sent: 27/11/2014 12:59:09
Subject: Re: GNU MediaGoblin

>On 11/27/2014 01:38 PM, Jessica Tallon wrote:
>>  Hey,
>Hi Jessica,
>
>>
>>  I've been working on the GNU MediaGoblin federation support which is
>>  designed to work with pump.io that uses Activity Streams 1. We're 
>>using
>>  a relational database and currently changing our current scheme
>>  (pre-federation) to a schema which will work with federated data. 
>>I've
>>  written up proposed schema we're going to be using in a document that
>>  largely consists of Python using SQLAlchemy models.
>Thanks for sharing it with the list! I hope we will see here more 
>emails
>like that soon :)
>
>>
>>  There are some legacy decisions such as retaining our internal 
>>integer
>>  primary keys in addition to "public" id's which are the federated 
>>ID's.
>>  The decision to keep these is partly due to the complexity of a
>>  migration from our current models and efficiency in PostgreSQL and/or
>>  SQLite (the two databases we currently support).
>* What do you plan to use as public IDs? HTTP URIs,
>acct:local@domain.tld style URIs ?
>* Will you internal IDs stay fully internal or you also plan to include
>them in JSON / HTML which you publish?
>* Do you already have some code which generates JSON / HTML from such 
>data?
>
>I noticed id, public_id, slug, username, url which on first sight seems
>to me like a lot of different types of identifiers...
>
>
>>
>>  Feedback on these models and possible things I've missed would be
>>  greatly appreciated.
>>
>>  https://theperplexingpariah.co.uk/GMG/new_models.html
>I'll take closer look at it in following days!
>
>Ciao o/
>

Received on Thursday, 27 November 2014 13:22:00 UTC