[Fwd: Re: [Linux-ha-dev] 2.1.1 is imminent]

[Lars Marowsky-Bree]

On 2007-07-11T23:44:45, Alan Robertson <alanr at unix.sh> wrote:

> Our friends at Novell put out a release which they named "2.0.8", which
> was not very much like Heartbeat version 2.0.8 (it had hundreds of
> changes and a few features not in 2.0.8), and which they unfortunately
> called "2.0.8", but which had only been tested by SUSE staff, and they
> put it out as service pack 1 of SLES10.
> 
> Although they did do relatively thorough testing on their hardware, it
> ran poorly on my cluster (had a number of regressions), and on our power
> PC cluster.

Alan, this is crap.

The 2.0.8-based version which we put out on SP1 was a _lot_ better than
2.0.8 of upstream at that time. No surprise, 2.0.8-upstream was 3-4
months old already, so that is expected.

If 2.0.8/SP1 ran poorly on your cluster, 2.0.8 as released was worse.

Our testing has more resources behind it then you do, we incorporated
bugfixes due to customer bugs which were found after 2.0.8 had been
released, so we _had_ to use a more recent version. 2.0.8 as-is was
unusable for our release as SP1; it would have been grossly negligent to
ship it at that time.

> Unfortunately, they didn't tell me they were going to do this, or it
> would have been better-tested and probably would have been numbered as
> 2.0.9 or 2.1.0. 

Uhm. Alan, compared to heartbeat, our SLE release roadmaps are pretty
long-term, at least 1 year into the future.

I won't bother to go back to the e-mail we've exchanged, but we've
spoken about our plans on the phone, I'm sure. I can't remember whether
you knew the exact date, but you definitely knew that we were working on
SP1. If you had really cared, you could have asked.

> I work hard to coordinate closely with them near their release times,
> but this time they declined to inform me of their plans.  They did
> tell some people in my company, but those people did not tell me, nor
> did Novell ask them to tell me.

Sorry, Alan. IBM and Novell share their product plans under NDA, and
you've always been privvy to them in the past. The only time I learned
that you were no longer because of IBM internal issues was after SP1 was
out.

And, do you think I kept/keep asking about a release plan and a roadmap
because I have nothing better to do? Why else could I've been interested
in in those, if not due to Novell's release cycles?

> This release had been labeled as "2.0.8", but was really much more than
> the real 2.0.8, and unfortunately not as thoroughly tested as I would
> have liked. 

This is crap. Your testing isn't better than ours. We've got 2,5 people
full-time on this project as developers; when we plan to release an
update, it gets hit by QA and pilot customers.

And, we pay a lot more constant and consistent attention to the project
than you do. Which is no surprise, because it's our full-time job, and
it'd be my guess Novell has a couple more customers paying for support
than IBM.

I have to call that bluff, because I will _not_ have our customers have
any reason to believe that our versions aren't sufficiently tested.
That's more or less my personal responsibility, so I get somewhat irked
at the suggestion that it might not be true.

> And, it had new features in it.  So, it _should_ have been numbered as
> 2.1.0.

As far as I remember, we couldn't agree on the version numbering scheme
at that time.

Andrew offered to make that release, but I seem to be recall that you
never gave him the final go ahead to do that.

> So, to try and get things kind of back in sync, we came to an agreement
> to skip that release in the numbering system, and the next release will
> be numbered as 2.1.1.
> 
> After this release gets finalized, for historical reasons I'll probably
> go back and put out a release called 2.1.0, which will have exactly the
> same content as Novell's SP1 "2.0.8".

Why would that be useful for anyone?

> I would have done it before now, since by definition, I can't change
> anything in it if it's to be a mirror of the SLES10 SP1 "2.0.8" version
> so I don't have to test it, but that would have delayed the 2.1.1
> release even more.

This is _so_ a redherring. It took so long because you didn't get your
act together, AND insisted (by passively blocking) on not delegating
anything. You didn't even reply to the mails I sent on that topic, some
even Cc'ed to the list.

> Although some others have implied that this new release is a surprise to
> them, it has been under discussion for some time, and we've been trying
> to put the final touches on it for the last few weeks, as the person in
> question well knows.

Alan, I've been asking you about your roadmap and plans for _YEARS_.
You've ignored all such requests, and now suddenly due to some positive
events, you start caring again, which will hopefully be sufficient
motivation and last until you get the next release out.

You've been telling me that 2.1.1 is "out soon" for a few months, so
eventually I stopped believing, and decoupled our releases from yours,
because they are too infrequent and too unreliable.

And, if you're talking about me, call me Lars.

> But, unless I throw down the gauntlet (and potentially ruffle a few
> feathers), a well-tested release will _never_ get out.  And right now,
> the release looks reasonable, better than the pseudo-"2.0.8" that was
> shipped earlier.  [But looking at test results indicates it still may
> have some regressions.  We'll see how major they are...].

_REGRESSIONS_!?!? "A well-tested release"!? "pseudo-"?!

You could almost have picked any version in the last 4 months and be
better than 2.0.8-upstream. It's to be expected. It's why we keep
developing it. That no "official" release went out for 6 months is a
shame for the project, and a nuisance for the users.

If you dare claim that our 'pseudo-2.0.8' had regressions, how many more
did 2.0.8-upstream at that time, if we had just taken it and shipped
it?

_Of course_ the current code looks better than our 2.0.8-SP1 release.
We've put in about 3 months of more development into it, and we're
working towards releasing it as an update to SP1 again, after it meets
our QA and feature requirements.

> And, of course, there's always the next release.  Which if we get this
> one out now, will come sooner than if we don't.

Oh, do you already have a tentative schedule? That would be a first.

> A few weeks ago a person on the project insisted that I was killing the
> project by not putting out a release RIGHT NOW even though the current
> version had known regressions.  Earlier today, after we fixed the
> regressions that were problematic then, this same person comes out
> whining that I'm putting it out too soon.
> 
> You just can't please some people.

Again, if you're talking about me, use my name, Alan.

The version of a few weeks ago was a lot better than 2.0.8-upstream. The
version we have right now is still better. There'll always be known
regressions and bugs. A well-tested release doesn't mean it's bug free.

I'm not complaining that you finally care enough again to put out a
release, I'm very happy about that.

I'm somewhat annoyed at how you try to make yourself look good and try
to blame it on us, and that you _never_ replied to the various mails on
that topic during the last few weeks. I could go back to point out that
you already sabotaged my attempts at a roadmap for putting out 2.0.0, or
all the various other run-ins we've had in the past, and where you keep
ignoring something and then eventually jump in and need it done "right
now".

I take it back. You _are_ consistent in the way how you "lead" the
project.

Luckily, it doesn't really affect our customers nor the people using the
packages from the build service, but it is a shame and an annoyance to
explain on the lists.


Regards,
    Lars

-- 
Teamlead Kernel, SuSE Labs, Research and Development
SUSE LINUX Products GmbH, GF: Markus Rex, HRB 16746 (AG Nürnberg)
"Experience is the name everyone gives to their mistakes." -- Oscar Wilde