Head stability [Was Re: [Fwd: Re: [PATCH] Fix Name mangling in HEAD]]

Andrew Bartlett abartlet at pcug.org.au
Tue Mar 26 04:31:11 GMT 2002

David Lee wrote:
> On 26 Mar 2002, Simo Sorce wrote:
> > [...]
> > Please Andrew, HEAD is an alpha version, if we do not experiment there
> > where shoud we? If I give an option (that was there in the first
> > commits) who would have used it?
> I'm suddenly getting worried...
> I fully appreciate that HEAD is developmental, and so non-production.
> Thus any of us "end-users" who choose to checkout from it do so on a
> strictly "own risk" basis, and that we should not put it anywhere near
> production.  Fine.
> Nevertheless, I had been under the impression, subject to the above
> caveats, that HEAD was basically intended to be heading towards stability,
> whose innovation aspects would be reasonably tested.  So that, although
> there might be temporary glitches, problems and inconsistencies, it would
> still be reasonably useable by those of us own-risk, read-only, end-user,
> development folk.

That is certainly the basis on which I suggest people give HEAD a try.  

> But Simo's comment implies that HEAD may be a place for experiment by the
> Samba Team.  Where does "reasonably tested innovation" end, and
> "experiment" begin?  I would have thought that "experiment" basically
> belongs in a non-HEAD branch or personal sandpit (just as any of us
> non-Team folk would do), with HEAD being reserved for stuff that has
> already basically had its principles peer-reviewed.

I certainly work on the basis that any code I commit to HEAD should
work, and should not have known issues that impact on default
configurations.  I will occasionally commit 'experimental' stuff into
code-paths that require explicit configuration, but I always aim to keep
the same core functionality.  

HEAD commits are generally not peer-reviewed (nobody really has the
time) but in general they are consistent, incremental changes that
doesn't impinge in existing functionality.  

This is partly why I maintain the Samba build farm - to test that we
don't break existing functionality, as well as keeping track of the
mundane issues like cross-platform compilation.

I certainly try to keep my random ideas in my personal CVS checkout
until I'm happy with how they work, or I get sick of the list of changes
and I scrap them.  In the past we have created CVS branches to allow
individual developers to peruse more risky changes outside the
constraints of HEAD.  (SAMBA_TNG is the most notable of these, but there
have been others).

> My own interest?  In the next couple of months we are making major changes
> to our local fileserving and authentication services, which will
> inexorably push us towards Active-Directory/PAM use of Samba.  So I was
> just about to check out HEAD for our local development environment.  If
> not HEAD, then where should I go for a stable-ish version of the emerging
> ADS/PAM Samba?

Likewise, I run HEAD in production (because I need some of its advanced
functionality).  Name mangling has been a running sore for a number of
weeks at my site, and I only just got time to start fixing it.   Other
than that, HEAD is remarkably stable and functional, and I want to keep
it that way.

Andrew Bartlett

Andrew Bartlett                                 abartlet at pcug.org.au
Manager, Authentication Subsystems, Samba Team  abartlet at samba.org
Student Network Administrator, Hawker College   abartlet at hawkerc.net
http://samba.org     http://build.samba.org     http://hawkerc.net

More information about the samba-technical mailing list