Technical Overview of the Alfresco / Jive Toolkit
Recently I transitioned from my long-standing role leading Alfresco’s Professional Services team to being the in-house technologist for the Business Development team, and one of my first tasks in the new role has been to work on an integration between Alfresco and Jive Engage. This work is being done in partnership with SolutionSet (a partner of both Alfresco and Jive), and I wanted to discuss some of the technical design work that has gone into the Toolkit, ahead of its availability (which will be soon after the Jive 5.0 launch – the version of Jive that the Toolkit is targeting).
Functional Overview, aka “What will it do?”
As announced at Gartner’s Portals, Collaboration & Content Summit this week, the integration (known as the “Jive Toolkit”) is a set of pre-built components that allows Jive to store documents in Alfresco while still offering all of the same social features as “native” Jive documents (commenting, rating, discussions, etc.). While not yet all-encompassing – Jive’s “social” content cannot yet be stored or managed within Alfresco – the Toolkit will provide a foundational level of document-centric integration, allowing implementers to focus on more use-case specific integrations as required (hence the positioning as a “toolkit”, rather than a fully fledged solution).
More specifically, the initial version of the Toolkit will allow users of Alfresco and/or Jive to create “managed” documents in any of the following 3 ways:
- By uploading a document to Alfresco, using the Jive UI.
- By “publishing” an existing document from Alfresco to Jive, using Alfresco’s Share UI.
- By “linking” an existing document stored in Alfresco to Jive, using the Jive UI.
In all 3 cases, the result is the same: the document is visible and accessible via the Jive UI in exactly the same way as any “native” document, but the content of the document is stored and managed in Alfresco only. Jive will maintain some metadata about the document – for example the document’s filename and a pointer to the document in Alfresco – but it will not store the binary content of the document. This approach ensures that the document is a first class citizen in both the Alfresco and Jive worlds, while minimising the risk of synchronisation issues between the two systems.
Here are some screenshots that demonstrate uploading a document to Alfresco using the Jive UI:
Technical Details, aka “Rubber, meet road”
As mentioned above, there are a variety of ways that the initial “linkage” of a document between Alfresco and Jive can be achieved, however all 3 creation mechanisms produce the same end state: Alfresco has the document in its entirety (including the filename, content, etc.) while Jive has a “proxy object” (a structured data-only object that has the filename and a pointer to the document in Alfresco, but does not have the actual binary content).
This means that all downstream events (updates, metadata modifications, deletes) can be handled the same way, irrespective of how the content was linked between the two systems in the first place – a major simplification in the logic for those downstream events.
Integration Mechanism, aka “CMIS, by any other name would smell as sweet…”
Another nice characteristic of this approach is that the calls from Jive to Alfresco (to create content, update and retrieve it) can be accomplished using the CMIS API. This has several benefits, from reduced development effort in the Toolkit itself (due to the ready availability of client-side CMIS libraries), to the potential for portability to other CMIS compliant repositories in the future.
One important thing to note is that the Alfresco-to-Jive API calls are not standards-based – they make use of Jive’s proprietary REST API. Jive does not expose a standards-based API (indeed, no suitable standard exists for social business systems yet), and CMIS doesn’t provide any kind of callback mechanism for clients to be notified when repository events of interest occur (i.e. a mechanism equivalent to Alfresco’s Component Policies).
Tricky Bits, aka “The Devil is in the Details”
As with any integration between complex enterprise applications, there is some trickery in some parts of the integration, and it’s critical to understand these if you’re evaluating the Jive Toolkit.
The first piece of trickery involves deletion of the content, specifically deletion in Alfresco. Because Jive maintains a pointer to the document in Alfresco (specifically, the “cmis:id”), rather than the content itself, if the document is deleted in Alfresco without Jive being notified, attempts within Jive to retrieve that content will fail. To prevent this, the Toolkit is currently designed to veto deletes in Alfresco if the document has been socialised in Jive. To delete a document, it will first need to be deleted in Jive at which point it can be deleted from Alfresco too. The reason the Toolkit doesn’t simply synchronise deletes between Jive and Alfresco is that there are common use cases where the document may be removed from Jive, but needs to be retained in Alfresco – replicating deletes between the two systems would have ruled out these use cases.
Full Text Indexing
The second item of trickery revolves around full text indexing of the document in Jive. To accomplish this, Jive will retain a copy of the content of the document just long enough to index it into Jive’s full text index, and once indexing is complete the content of the document will be removed from Jive. As you’d expect, Alfresco will also notify Jive of any updates to the document, so that the content can be re-indexed on the Jive side.
Access Control and Identity
Access control to the documents is also tricky, primarily because the Alfresco and Jive ACL models differ in their level of granularity. Jive’s access control is primarily Community-centric (i.e. defined and enforced at the level of the Community), while Alfresco has a fine grained, per-node (file or folder) ACL mechanism. In this first release, the Toolkit will initially create the document in both systems in such a way that the ACLs are in sync, but modification of those ACLs in either system will not be replicated to the other system. The upshot is that direct manipulation of the document’s ACLs in Alfresco may cause errors in Jive (i.e. users who can see the document in the Jive UI, but are unable to download it).
Furthermore, in order for Alfresco and Jive to agree on the principal set, the initial version of the Toolkit assumes that both Alfresco and Jive are configured to use the same LDAP repository for user identity and authentication. During the design sessions it was felt that this was likely to be a requirement for an integrated solution anyway and hence wouldn’t be an impediment, but we’re keen to have that assumption validated as broadly as possible.
So there you have it – a whirlwind tour of the upcoming Jive Toolkit! As a v1.0 there are some more sophisticated use cases that the Toolkit doesn’t address yet, including multi-document / library based integration, and capture of Jive’s social content (discussions, ratings, wiki pages, etc.) in Alfresco. The intention with the Toolkit is to initially provide Alfresco+Jive Systems Integrators (such as SolutionSet) with a small but solid base on which such extensions could be built, and if/when common requirements are identified for these more sophisticated use cases they can be rolled back into the Toolkit.
We’re keen to hear your feedback and look forward to your participation in the project!