[erlang-questions] FOP (was: Re: Trace-Driven Development)

[Michael Turner]

On Fri, Jun 8, 2012 at 6:55 PM, Ulf Wiger <ulf@REDACTED> wrote:
>
> On 8 Jun 2012, at 09:19, Michael Turner wrote:
>
>> If I submitted a patch to the seq_trace documentation so that it
>> mentioned that seq_trace implements Lamport clocks, and it got
>> rejected with "we're still thinking about how to work that fact in,
>> maybe we'll have something in the next release," I think I'd end up
>> writing Lamport to say, "Can you *believe* these people?"
>
> Up to this point, I thought you were on a roll with clear,
> insightful and respectful discussion not least about the
> problems of understanding each other in writing, and
> conveying ideas through documentation.
>
> I can understand that certain points can get lost in a long
> and heated thread, so let me clarify a few things:
>
> - I don't work for Ericsson (as also apparent from my sig)

Ulf, at no point in that particular paragraph did I mention you by
name. We didn't exactly bury the hatchet, so it's not like you're
digging it up. But you are picking it up.

> - I don't vet patches on OTP. [....]

I never suggested you did.

> - Whatever changes *I* would like to see to seq_trace, are
>  my own ideas; [...]

Nothing I wrote up there about getting seq_trace documentation to
mention that it implement Lamport clocks could reasonably be construed
as your own idea, particularly since you initially seemed pretty hazy
in your response my statement that seq_trace implements Lamport
clocks.

Specifically:
"... many of the inputs that *actually* informed the implementation
were either confidential or proprietary enough to be of no interest to
the people reading the manuals."

You stressed "*actually*". To me, this strongly implies that Lamport
clocks -- whether they came in as a recommendation by Erlang/OTP as a
"better forlopp tracing", or because of a customers request for them
(in the other version of two as-yet-unreconciled versions of events)
-- did not "inform the implementation." Are they in there? Then there
was some kind of "accident." Or "coincidence." Which seems ridiculous,
since it's pretty obvious from citations in the open literature by
half a dozen Erlang/OTP people that they already knew what Lamport
clocks were.

> In this particular case, the most relevant passage of our
> long discussion should be Gustav Simonsson (who *does*
> represent OTP) saying that your patch was welcome:

If you mean specifically the idea of a patch mentioning that seq_trace
implements Lamport clocks, I have yet to see that from Gustav.

He wrote that documentation patches are (generally) welcome.
Obviously, that's no guarantee of the acceptability for any particular
one.

And I have it (second-hand from you) that he has himself thought of a
good place for this mention, in the seq_trace documentation. To which
I replied: if he's got a good idea for that, why doesn't he just do
it?

I mean, c'mon, this is on the web. It's not like you have to do a
whole print-run of some dead-tree reference manuals to fix it today.
And every day that the Lamport clock pseudocode sits there in the
seq_trace documentation is one more day when Erlang/OTP looks like
it's plagiarizing the idea of a major computer scientist.

As far as I can tell, NOBODY at Ericsson has, in writing, said
directly, "We need to fix that." Until then ....

> The only concrete comments you've gotten this
> time from OTP have been words of encouragement: submit
> a patch, and it will be accepted as long as it improves the
> documentation - not a single hint that they are planning
> anything else for seq_trace.

And I've already responded to that: "improvement" is in the eye (and
to some extent, the imagination) of the beholder. So long as it's part
of some internal Ericsson mythology that seq_trace only reiterates
Lamport clocks as a technique invented independently (and
concurrently, if not earlier) in "forlopp tracing", the response can
always be: "we invented that, probably before Lamport thought of it,
we don't have to credit anybody else, and some of our customers still
think we invented it anyway, so leaving it unmentioned is what works
best for Ericsson AS A BUSINESS."

Until I hear directly from Erlang/OTP people otherwise, I have to keep
that possibility open in my mind. Especially since both you and Robert
Virding seem to have been under this impression, if you aren't still.

> We have no reason to think
> that they are, but now would of course be a good time
> to say so then. :)

I'm all ears.

> There are others who are actually on the fence: Joe,
> Richard Carlsson, Kostis, … who are not on the OTP
> team, but deeply enough involved that it can
> sometimes be hard to make out whether they speak
> for the Team or not. In case of doubt, it's best to ask
> for clarification and generally assume they don't.

I hope it doesn't have to go to lawyers for approval. I mean, whether
forlopp tracing, at any point in its evolution, did or did not include
Lamport clocks ...  is that actually something you can't reveal under
NDA? The mind boggles. If it's in there, the fact that it's also a
technique in the open literature from around the same time would seem
to argue that there's no big deal (in NDA terms, anyway) in telling
people. Likewise if it's *not* in there -- you don't even have to
divulge details of what IS in there -- you just issue a denial.

I can think of reasons why a corporation might be cagey on an issue
like this. None of them are terribly complimentary to the corporation.

-michael