[rsyslog] rsyslog version numbering
rgerhards at hq.adiscon.com
Sun Mar 30 11:01:22 CEST 2008
[Mmmhhh. I thought I had sent a reply from the pda. Well, please don't
wonder if it shows up some time later ;) So here is the full response:]
Any version with a devstate designation (-mf, -rc) is unstable. Every
version without it is stable (this will be far less versions as we reach
stable only every now and then).
So what would the current version be named? Mhh... It's not an easy
one-to-one mapping. Because the process changes a bit. Today, I add new
major and minor features at the same time while stabilizing old
features. This is also the primary reason why we so far have very
infrequently new features in the stable release: whenever I stabilize an
"older" feature, I introduce new bugs while working on the new ones.
That'll change a bit with the new scheme, as I then will freeze code
once the focus feature has been implemented. When I start with the next
focus feature, patches will be applied to both the "older" code as well
as the one I am working on. As such, features become stable during the
process. It obviously is some more work to do for me, as I now need to
apply patches to multiple code paths (to keep this reasonable, I limited
the stable releases to only the latest minor version). The plus is that
we get much more feature-rich stable releases, so I think the additional
development time is well spent. And for the ultra-conservative folks,
there is always the old, feature-less previous version stable (v2 in
After having said this, the current release would probably have two
3.12.0-rc3 --> the stabilized module loader release
3.13.0-mf2 --> the new relp-enabled release which has not yet real relp
I am also now of the thought that we do not necessarily need to move to
4.0.0 to cover this new scheme. I'd say I can simply release a 3.13.0
stable next week (based on the current 3.12.5) and then continue to work
on relp in 3.14.0-rc0 (there will probably be no -mf for that release as
relp support is already quite far).
Question now: how can I just declare the release to be stable even
though it contains the RELP code? Answer: I am lucky ;) the relp code
*inside* rsyslog is just two slim inout/output plugins. No core changes.
The actual relp code comes via librelp, which is external. Thanks to
this coincidence, I did not really introduce any new features since the
last feature focus and so the rest of rsyslog could mature. At no other
point in the v3 tree I would have been able to do this. That is yet
another reason that I'd like to settle the issue by mid next week - the
initial version of relp is almost done and I am eager to move on.
Again, feedback appreciated.
> -----Original Message-----
> From: rsyslog-bounces at lists.adiscon.com [mailto:rsyslog-
> bounces at lists.adiscon.com] On Behalf Of Michael Biebl
> Sent: Saturday, March 29, 2008 3:16 PM
> To: rsyslog-users
> Subject: Re: [rsyslog] rsyslog version numbering
> 2008/3/29, Michael Biebl <mbiebl at gmail.com>:
> > 2008/3/29, Rainer Gerhards <rgerhards at hq.adiscon.com>:
> > > I needed some time to digest all this myself ;) I've now done that
> > > put it all into a HTML page. That doesn't imply it is the scheme
> we will
> > > finally use, but if so, it saves me some time. I have also
> described the
> > > development process a bit. I think this is vital for
> understanding why I
> > > need certain release numbers. The info is here (with the now
> > > text at the end of the doc):
> > >
> > > http://www.rsyslog.com/doc-version_naming.html
> > >
> > > I am still a bit undecided if we really need to go v4 or can
> start this
> > > (or whatever else) in the v3 branch...
> > >
> > > Feedback appreciated.
> > What I'm missing from the document, is a distinction between stable
> > and unstable releases
> > (or are all releases with -rc? unstable and without stable?)
> I.e., are -mf releases to be considered stable or not?
> Would the current 3.12.5 be an rc release in the new versioning
> Why is it that all of the instruments seeking intelligent life in the
> universe are pointed away from Earth?
> rsyslog mailing list
More information about the rsyslog