[Vm-dev] explanations

Thu Apr 5 14:51:05 UTC 2018

and this is my reply, again posted with Esteban's permission.

Hi Esteban,

> On Apr 4, 2018, at 4:42 AM, Esteban Lorenzano <estebanlm at gmail.com> wrote:
> 
> Hi Eliot, 
> 
> Sorry I was not around yesterday when you pinged me. 
> I will try to explain here why I removed myself from OSVM.
> 
> First, this has nothing to do with my general appreciation of you and your work. You know I consider you my friend and of course I valour your work.

And you know I feel the same way about you.  That makes this doubly disappointing.

> But (there is always a “but”): this is not working for Pharo and me.
> 
> I could point several things I consider problematic:
> 
> - the lack of a solid process that we may feel confortable.
> - the conservative approach to change things.
> - lack of communication.
> - the “de facto” stagnation of most proposals (some times even trivial) that comes from Pharo world and the consequent degradation we see on pharo stability.
> 
> I think most of this problems come because there is a different culture that creates an impedance between the two ways of work.  Also, there is a subjacent rejection of Pharo that emerges as part of the culture of the vm-dev community that makes us not feel welcomed, nor our suggestions. And I’m not the only one to feel it, I’m just the last one to react.
> I know, each time something like this is told, everybody "tear their clothes” saying that’s not true, but actions talk better than words and that’s the general (negative) feeling of people coming from Pharo trying to do VM.
> 
> And after seeing that the only response in list was an olympic “is all your fault, deal with it”, I decided it was enough for me.

Was that my message pointing out that the Pharo vm was being built from old VMMaker source and that the bug was already fixed?  That was just a fact.  I'm very sorry I was not very diplomatic but the fact remains be that the vm /was/ being built with generated sources that were months out of date.  I don't understand why that is.  The Squeak vm had the fix already.

> So yes, I’m frustrated. I made a lot of efforts to come back from the fork, and now I wonder if I made the correct step, since as we say in Spanish “mejor ser cabeza de ratón que cola de león” (better be the head of a mouse than the tale of a lion).
> 
> But I’m no troll so instead starting a flamewar (in a moment where you are receiving an unfair attack from a sick guy… an attack btw you have all my solidarity) I just preferred to fade out.

Let me gently suggest that you're wrong about participation.  You're not being ignored.  We're most of us volunteers, and while you get paid to work in the Pharo VM, I got paid to work on Newspeak and am losing my job in July.  The folks at HPI are working on their own vm and the SSL work bleeds over.

The counter example is Alistair Grant who did the work necessary to get changes to the FilePlugin that are for Pharo released in a few months.  His first code was wrong architecturally (the primitives didn't fail, they answered error codes, and so they broke the Sour forwarding scheme). I explained.  He rewrote; a lot of work.  I then improved primitive failure support so that he could answer OS error codes in primitive failures and fixed some slang issues.  He polished the code in response to my review.  Now we have a much improved FilePlugin.

Another counterexample is Clément's recent work with Sophie on new comparison and copy primitives.  They did the work, it got in.

So when you say
"the “de facto” stagnation of most proposals (some times even trivial) that comes from Pharo world and the consequent degradation we see on pharo stability." I think "I am expected to do the work".

And when you say "the conservative approach to change things." I can point to the Sista work, Spur itself, the new file plugin, the prompt fixes I made to the compactor as soon as I got a report from a Pharo user about the bug and so on.  We do change.  But you're probably talking about change in process.  That's more difficult because we try and work together and changing the process means getting lots of people and lots of code to change.  I didn't invent the process.  I didn't originate opensmalltalk/vm.  But I do see that we have a process and I stand behind it.  But I also see that it can be improved (I am trying the get somewhat be to rewrite the Linux build to use makefiles as we speak, for example).  But I'm not able to do the work.  Changing, or rather evolving the process, is something we have to do together.  And your leaving vm-dev doesn't help you in helping the community make the changes you want.

Now we have some real problems:
- the manual source generation step
- no issue tracker for the VMMaker side of the code base
- multiple tips (VMMaker (Sista, Spur), SSL, other plugins, platform C code, CI infrastructure, ...)
- a one-level repository/CI infrastructure that doesn't allow anyone to test independent of anyone else, so we always end up breaking each other's code.

I believe that the key problems are the multiple tips and the lack of a two-level structure to allow us to decouple the tips and test components independently and hence be able to compose the production vm from stable components, not from unstable tips.  Once that is possible source generation doesn't have to be manual (note that Ronie, Clément and Nicolas understand the source generation steps and have done it and committed themselves, so I am not the only one who does this).

I could be wrong.  And setting up a two level structure will be costly; every change yo VMMaker will have to be done twice.  The CI infrastructure will have to be much more complex and be maintained.  So this needs to be properly discussed.  And vm-dev (not Pharo dev as do many times is the case) is the right place to discuss it.

I wonder if you would be willing to post your message (this one I am replying to) to vm-dev and for me to post this reply (missing this last paragraph) to start a proper discussion about
- what is wrong with /our/ process
- what we can do to fix it
- who can do the work and in what timeframe

And I am very very sorry I have not made you feel welcome and included and heard.  I have failed you.

> cheers, 
> Esteban

cheers!
Eliot
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.squeakfoundation.org/pipermail/vm-dev/attachments/20180405/933ae1e8/attachment-0001.html>