Running SVN trunk vs a stable VICIdial release

VICIdial ships from Subversion, and you pick what you check out: the trunk, where new code lands daily, or a tagged release, frozen and tested. Both run the same dialer. The difference is risk. Trunk gives you the newest fixes before anyone else; a release gives you a known quantity that thousands of boxes already run. For a production call center, that choice is not academic.

What trunk actually is

Trunk is the live development line. When a developer commits a bug fix or a new feature, it goes to trunk first. Run svn update against trunk and you pull whatever was committed since your last update, including changes that have had little real-world testing. That is the appeal and the danger in one sentence. You get fixes early, and you get the occasional regression early too.

A tagged release is a snapshot of trunk at a moment the maintainers judged stable, frozen and given a version number. It stops moving. The bugs it has are known and documented; the surprises are mostly behind it. For a box that has to dial all day, predictable beats new.

flowchart LR
  A[VICIdial Subversion] --> B[Trunk live commits]
  A --> C[Tagged release frozen]
  B --> D[Newest fixes early]
  B --> E[Possible regressions]
  C --> F[Tested and predictable]
  C --> G[Fixes arrive later]
  D --> H[Staging box]
  E --> H
  F --> I[Production dialer]

When trunk earns its keep

There are real reasons to run trunk. You hit a bug that was fixed in a commit but not yet in any release. You need a brand-new feature for a campaign launch. You run a staging box specifically to shake out trunk before it reaches your production server. In each case the move is the same: pull trunk on a box that is not taking live calls, and watch it. The install script walkthrough shows how to stand up that second box.

What you should not do is run trunk on the only box you have and dial production traffic on it. A regression in the auto-dial path or the agent AGI (Asterisk Gateway Interface) handler can stop calls, and you find out when agents go idle. The cron-managed Keepalive will faithfully restart broken code, which does not help. One box, one campaign, no staging means run a release.

The pragmatic default

For most operators: run a stable release in production, and keep a staging box on trunk if you want early fixes. Pin your production server to a known revision and upgrade it deliberately, on a snapshot, rather than tracking trunk's daily churn. That keeps the Asterisk layer and the VICIdial layer both predictable, and it keeps your Dialer pacing behaving the same way today as it did yesterday. Because each box is Single tenant, your revision choice is yours alone, with no shared platform to coordinate.

For where this sits in the install lifecycle, see the installation guide, which covers picking a revision during bring-up.

On our managed boxes we ship a tested, secured build and patch Asterisk CVEs for you, so you are not exposed to trunk's rough edges by default. You keep root SSH, so if you want to track trunk or pin a specific release on a VPS, you can. Provisioning takes under 40 seconds, which makes spinning up a throwaway staging box cheap. See plans and pricing.