f1xska

You've loaded an old revision of the document! If you save it, you will create a new version with this data.

<h1>F1x Architecture</h1>
<div class="level1">

</div>

<h2>Overview</h2>
<div class="level2">

</div>

<h3>Terminology</h3>
<div class="level3">
<ul>
<li class="level1"> Site - F1x node installation</li>
<li class="level1"> Identity - public/private keypair</li>
<li class="level1"> RealmID - hash of Identity public key</li>
<li class="level1"> Realm - set of objects that are related to an Identity. All shards in realm have the same realm ID. <br/>
Likely to be implemented as git repository, for its strong builtin sync and reconciliation feautures.</li>
<li class="level1 node"> Object - contains fields of Object or Code (Handler) kind with unique name ~ObjectID<ul>
<li class="level3"> mandatory fields: name, kind</li>
<li class="level3"> Data shards have _handler</li>
<li class="level3"> Code shards have _code attribute</li>
<li class="level3"> Handler PubSub has method listener_SOMENAME and callback_SOMENAME. The callback method of Handler is called if result of the corresponding listener method changed. Internal logic might intentionally call listener method to invoke subscribed callbacks.</li>
<li class="level3"> version - used by reflectors and Router to facilitate subscriptions</li>
</ul>
</li>
<li class="level1"> Shard - contains specific version of ObjectID kind with unique address ~ShardID</li>
<li class="level1"> Handler - Code shard that knows to execute methods for certain Object kind</li>
<li class="level1"> Handlers Set - Handlers published by the same Realm</li>
<li class="level1"> Reflector - Object sole purpose is to subscribe to a listener method of referenced PRIVATE Object. Has unique ReflectorID (likely amalgam of source ObjectID). On every change - new public shard (with updated version) is created, with the new result value is stored, the shard is signed and made available for external subscribers. <br/>
For security reasons, normal and published Shards have different location on disk.</li>
<li class="level1"> Publish - procedure to create Reflector</li>
<li class="level1 node"> Subscription - Object of kind Subscription that specifies rules to invoke methods of other Objects.<ul>
<li class="level3"> Subscription to Reflector: specifies ObjectID→method to call on each new version of specific ReflectorID</li>
<li class="level3"> Subscription to Objects: specifies ObjectID→method to call on each new Object of kind "X" and possibly other conditions (like geo "Y", gender "G" or age range MIN/MAX) - used by <a href="/doku.php?id=f1x:archtecture:attraction-proximity" class="wikilink1" title="f1x:archtecture:attraction-proximity" data-wiki-id="f1x:archtecture:attraction-proximity">Attraction/Proximity</a></li>
</ul>
</li>
<li class="level1 node"> Router - Caching/Routing/Matching Server (CRMS) - intermediate server, usually sits on internal network or at ISP<ul>
<li class="level3"> Router - keep track of objects its children sites are subscribed to, forward objects to sites. Allows Sites to register as children based on IP.</li>
<li class="level3"> Cacher - cache frequently accessed objects</li>
<li class="level3"> Matcher - <a href="/doku.php?id=f1x:archtecture:attraction-proximity" class="wikilink1" title="f1x:archtecture:attraction-proximity" data-wiki-id="f1x:archtecture:attraction-proximity">Attraction/Proximity</a> role</li>
</ul>
</li>
</ul>

</div>

<p>
Site is compute infrastructure that enables users to operate Object Realms. It consists of events routing server and presentation (like web server). <br/>
 <br/>
Web server (presentation) functionality:
</p>
<ul>
<li class="level1"> authenticate the site user (a site normally has one user, preferably with unique domain)</li>
<li class="level1"> communcate with core on objects CRUD</li>
<li class="level1"> visualise and manage content in user-friendly way</li>
<li class="level1"> show notifications</li>
</ul>

<p>
Events routing server performs functions below:
</p>
<ul>
<li class="level1"> Storage for shards</li>
<li class="level1"> Execute methods</li>
<li class="level1"> provide pserver interface for exported Shards</li>
<li class="level1"> Keep track of subscribers, notify CRMS on changes</li>
<li class="level1"> garbage collect "foreign" Shards that are not referenced anymore</li>
</ul>

</div>

<h4>Data Shard</h4>
<div class="level4">

<p>
Data Shard is on-disk representation of object with fields that has associated data handler (Code Shard), similarily to data and code separation in <a href="https://en.wikipedia.org/wiki/Object-oriented_programming" class="urlextern" title="https://en.wikipedia.org/wiki/Object-oriented_programming" rel="ugc nofollow">OOP</a>. <br/>

Code Shard methods is written in <a href="https://en.wikipedia.org/wiki/Lua_(programming_language)" class="urlextern" title="https://en.wikipedia.org/wiki/Lua_(programming_language)" rel="ugc nofollow">Lua</a> and is executed in sandbox. API methods to read and call referenced Objects allows writing sophisticated applications.
</p>

</div>

<h3>Published Shards</h3>
<div class="level3">

<p>
Normal shards are accessible to Realm owner only. The owner might choose to publish methods of some objects using special "Reflector" objects. These also might be encrypted with public key of intended recipients, thus limiting who can see the content.
</p>

</div>

<h2>pCache server</h2>
<div class="level2">

<p>
Every Site is connected to a pCache server. It might be discovered by asking centralized "pCache servers registry". It is publicly accessible and should be at location with good connectivity. The pCache servers perform functions below:
</p>
<ul>
<li class="level1"> retrieve list of subscriptions from sites under its control, send back matching Memos</li>
<li class="level1"> retrieve "exported" data from sites and share it to sites that request it. Cache the data.</li>
<li class="level1"> receive notifications from children sites about data change, forward them to upstream pServers or other sites</li>
</ul>

</div>

<h3>Publish/Subscribe</h3>
<div class="level3">

<p>
Frequently object want to know when something changes in other specific objects. To maintain encapsulation principle, this is done by "listener" methods. When other objects indicate desire to get notifications of a "listener" method, result of the method is remembered, together with list of "subscribed" object/methods. When Object wants to "notify" subscribers on a change, it invokes listener method that returns different value, thus causing methods of subscriber objects to be called.
</p>

<p>
The subscribers might then recalculate a value, compare with threshold and raise nofification to user or alter return value of their own listeners, thus "publishing" a change.
</p>

</div>

<h3>Attraction/Proximity</h3>
<div class="level3">

<p>
To discover other objects of same kind, there is <a href="/doku.php?id=f1x:archtecture:attraction-proximity" class="wikilink1" title="f1x:archtecture:attraction-proximity" data-wiki-id="f1x:archtecture:attraction-proximity">attraction-proximity</a> "wandering search query" mechanism.
</p>

</div>

<h4>Task "match riders that go from Jerusalem to Eilat"</h4>
<div class="level4">

<p>
Objective: My Object "Intend to travel from Jerusalem to Eilat by car, 2 empty seats" need to find other objects "Intend to travel from Jerusalem to Eilat somehow". TBD
</p>

</div>

<h3>JQ queries</h3>
<div class="level3">

<p>
Microapps need to access data from other Memos. They can run <a href="https://jqlang.github.io/jq/" class="urlextern" title="https://jqlang.github.io/jq/" rel="ugc nofollow">jq</a> queries on "known" Memos:
</p>
<ul>
<li class="level1"> what is default measurement system of the owner? Imperial or standard</li>
<li class="level1 node"> get list of neighbours who know me:<ul>
<li class="level3"> find Memos of kind People and have homeAt geo and have Memo of kind Subscription where (site=me and kind Memo=Message)</li>
<li class="level3"> remove ones that have distance between me and homeAt more than 100 m</li>
</ul>
</li>
</ul>

<p>
Stored queries (subscriptions) also use jq to find matching Memos: TBD How?
</p>

</div>

<h2>Security</h2>
<div class="level2">

<p>
For the system to be popular, users must trust the system to be able to protect their data. That means that Handlers must be trusted not to misuse data.
</p>

<p>
During setup Site owner is assigned "default" set of Handlers, that been carefully audited. He might change it or "overlay" it with other Handlers sets.
</p>

<p>
In addition to trusted Handlers set, data sharing must be permitted for each case. Initially Handler has "Unknown" security level. Once it reads from internal object, it is assumed "restricted" level, and it can not publish (write into public objects), unless permitted by Site Owner. The allowance might have a checkbox "remember this choice", thus allowing the particular combination of Handler and Object to publish.
</p>

</div>

<h3>Site security</h3>
<div class="level3">

<p>
Handlers are executed in sandbox, with access to Object fields (in the shard). Some fields are volatile, e.g. they are stored in cache area, separately from the shard. For POC we use <a href="https://www.php.net/manual/en/class.luasandbox.php" class="urlextern" title="https://www.php.net/manual/en/class.luasandbox.php" rel="ugc nofollow">Lua</a> for Handlers language (<a href="https://pypi.org/project/RestrictedPython/" class="urlextern" title="https://pypi.org/project/RestrictedPython/" rel="ugc nofollow">RestrictedPython</a> is possible).
</p>

</div>

Enable Complex Tables Editor Height px Turn supplementary image paste support on

Edit summary: Minor changes