[OOTB-hive] WG: [ADDONS] Proposed activity details

Bindu Wavell bindu at wavell.net
Thu Aug 21 01:43:06 BST 2014


I'd love to help define and support best practices for addons. I am opinionated about this (also open minded) have discussed this a bunch with Alfresco relative to support-ability and also internally at Zia as we are doing customer work. Which committee would that bee?  +1 for Alfresco BEE = BEe Edition or BEE edition ;)    

-- Bindu

On August 20, 2014 at 2:57:35 PM MDT, Martin Cosgrave <martin at ocretail.com> wrote:   On 20/08/14 00:48, Axel Faust wrote:       Activity #1: Form list of Order-reviewed and –accepted Alfresco add-ons    ·         Non-commercial / open-source add-ons only (anyone – ourselves included – should be able to use it / modify it)       essential      ·         Add-ons listed on addons.alfresco.com only (this is and should remain the central directory)       +1      ·         Definition of detailed eligibility / acceptance criteria    o   Eligibility: Do we only include add-ons in the narrow sense of extension of Repository and Share, or tools / scripts for administration and/or development as well?       The wider sense, otherwise how do we include the Loftux install script? ;-)      ·         Successful compatibility test on most recent / most stable Alfresco CE version(s) (currently Alfresco 4.2f + Alfresco 5.0a - may be a single version in the future)       I was thinking about this. While it would be lovely to track the latest and greatest, I feel strongly that we should not, but instead track the most stable, in this case 4.2.f. We will simply not have, and probably never will have the resources to track the bugs on the new releases in the way that alfresco can.      ·         Successful compatibility test with other add-ons already reviewed / accepted   ·         Successful compatibility test with current supported client environment       Hopefully we can have a one-time setup where the addon submitter verifies their addon with our test systems, and gives us a test suite which passes in that environment. To be included they would need to sign up with us I think, and if they didn't respond to our technical questions, possibly a couple of years in the future, they would get removed. We should track the bugs and assign the fixes to them (if we can, depends on how important we are or they are!) and expect to get tests added to the suite when a bugfix is applied, which demonstrate the fix.      ·         Basic functionality test from user perspective (using available documentation)       Selenium scripts for functionality tests would be something to aspire to.      ·         General technical review    o   Potential security issues (unsecured services, improper runAs usage, lack of traceability)    o   Potential scalability issues    o   MT capability    o   Configurability    o   De-installability / de-installation procedure    o   Development best practices (details TBD, e.g. namespacing, support of separate Repo + Share + SOLR instances, use of public services, no hard override of Alfresco out-of-the-box files or beans …)       This seems like a whole separate track to me, and I want to be on it! "TECH", or do we have that already?          Rationale: Our list needs to provide a significant “added value” to complement addons.alfresco.com without being “just another top 10 list” that you can find on some blogs already. The majority of us should be able to say “I would have no qualms using that add-on in production” for every add-on we list. Our tests / review should never be a replacement for proper QA by the add-on developer, instead focusing more on the “Model Citizen” aspects, interoperability and compatibility. As a group, we should have enough experience to review add-ons concerning the more advanced concepts (MT, scalability, de-installation) which a single (beginner) developer may not yet be aware of but can hurt unaware add-on users the most…       That sounds like the disclaimer that we will need to have at the bottom of every "BE" addon page ;-)           I would to see us come up with staggered sets of acceptance criteria, e.g.   ·         Mandatory core criteria that all add-ons must pass unconditionally (e.g. CE compatibility, compatibility with at least X different clients, separate Repo + Share + SOLR instances, no hard overrides…)   ·         Secondary criteria with limited allowed violations (e.g. improper runAs usage, compatibility with other add-ons…)   ·         Tertiary criteria that will result in some sort of flagging, warning or simple documentation within our listing       Anything that can be automated should be prioritised IMHO. (Sidebar: what is "improper" runAs usage and how can we detect it? Remind me never to tell you about the time I extended the auth system so it didn't need ALF_TICKET any more ;-) I suspect there was lots of improper runAs in that project).  Static code analysis is a fascinating topic but somewhat beyond me. "Compatibility with other addons" might be more straight forward in a way, as long as we have many automated tests to be able to understand which addon is breaking things (i.e. automated builds of the system excluding each addon one by one).           Activity #2: (long term) Continuous update / maintenance of add-on list    ·         Trigger #1: Updated Alfresco CE release       I strongly feel we should only track the stable release, i.e. 4.2.f at the moment, maybe 5.0.e/f/g at some point? But maybe this is a cop out for our "customers" (who don't pay us :P) who would like the newer releases? We would have a lot less effort to expend if we just tracked the senior stable version though. Who knows, we may never move to 5.0, just extend, enhance and fix 4.2? Not saying that's a goal but it could happen. Also we could call it "BE 42" and Boriss would be ecstatic (he's a fan of Hitch Hiker's Guide to the Galaxy) ;-D       ·         Creation / Setup of tooling / test automation for basic compatibility / functionality tests    o   Prepared reference environment (e.g. Amazon Machine Image incl. reference data set + CloudFormation template)       I would love to standardise on Vagrant + puppet for this. It makes it super easy to open source the devops config that way.       o   Source / build project to aggregate all / subset of listed add-ons for testing    o   Individual UI test script / test suite for each add-on (Selenium or anything else (not my strong area))    o   Potential test script / test suite for backend services of each add-on       Probably my earlier cursory reading of this mail triggered my thoughts which I've already mentioned above about tests. WE NEED THEM! :D Also I have done a bit with Selenium so I am up for writing scripts and/or knowledge transfer.   ·         Flagging of listed add-ons that no longer match acceptance criteria       One of which is the developer responding to an occasional re-confirmation email.      ·         Dropping / removal of listed add-ons that repeatedly failed acceptance criteria / are no longer actively maintained       Everyone who contributes to the Honeycomb should know that bees always remove decaying matter from their hives. When the decaying matter is too big to remove (e.g. a mouse gets in and dies) they surround it with propolis which is like bee duct tape: http://en.wikipedia.org/wiki/Propolis   Buzz buzz Martin     
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.xtreamlab.net/pipermail/ootb-hive/attachments/20140820/d67c2c36/attachment-0001.html>


More information about the OOTB-hive mailing list