I've just updated hackydev with new stable releases for all configurations based upon last night's build:
<http://hackydev.ics.hawaii.edu/hackyDevSite/publicRelease.do>
The reason for notifying you of this is because I'm going to slightly change my approach to the notion of 'stable' release. Up until now, I basically maintained a release approach of: (a) there are nightly builds of unknown stability. (b) there is exactly one stable release per minor version number. (in other words, version 6.3 had one stable release, version 6.4 had one stable release, etc.) (c) each stable release is accompanied by an email to all of the mailing lists (which happens about every six to eight weeks).
The problem with this approach is that there are certain daily builds between stable releases that are more stable than the 'official' stable release, and I would like to be able to capture these intermediate builds.
So, my new approach is as follows: (a) there are still nightly builds of unknown (and frequently poor) stability. (b) there are _multiple_ stable releases per minor version number. Whenever the system reaches a state where the net level of bugs seems lower than the current stable release for a configuration (or where a really major bug has been identified and removed), I'll make that the new stable release. I'll send email to hackystat-dev-l when that happens, but not to hackystat-announce-l or hackystat-users-l. (c) I will send email to hackystat-announce-l and hackystat-users-l for the _last_ stable release of a version series, in other words when we've closed all the Jira issues scheduled for that release and are ready to increment the version number and start on a new round of instability. This email will include a summary of the new features and so forth incorporated in this version, just like always. This maintains the six to eight week update schedule for those folks.
The biggest problem with this approach that I can see is that it might be nice to be able to maintain access to "the final stable build associated with a version". We can, of course, always construct it from CVS given the date of the email to the mailing lists, but automatically maintaining it might be desirable in future.
Let me know if you see any other issues.
Cheers, Philip
