<div dir="ltr"><br><div class="gmail_extra"><br><div class="gmail_quote">2015-10-23 19:01 GMT+02:00 Eliot Miranda <span dir="ltr">&lt;<a href="mailto:eliot.miranda@gmail.com" target="_blank">eliot.miranda@gmail.com</a>&gt;</span>:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"> <br><div dir="ltr"><br><div class="gmail_extra"><br><div class="gmail_quote">On Thu, Oct 22, 2015 at 12:05 AM, stepharo <span dir="ltr">&lt;<a href="mailto:stepharo@free.fr" target="_blank">stepharo@free.fr</a>&gt;</span> wrote:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"> <br>
  
    
  
  <div bgcolor="#FFFFFF" text="#000000">
    +1 <br>
    (I&#39;m amazed that you can type all that on an iPhone). <br>
    <br>
    Eliot I hope that the blog and the lecture clement is designing will
    help people to get closer to the VM. <br>
    I would love that Inria hires a VM researcher and that the rmod
    group gets exposed and push the VM. It did not work with Stefan Marr<br>
    which could have got a position but was home sick :). <br>
    If you know people that have a good cv and want to live in France,
    we can fight to get a permanent life researcher position. <br></div></blockquote><div><br></div><div>Good to hear!  I&#39;ll keep my eyes peeled, as they say :-)</div></div></div></div></blockquote><div><br></div><div>But please, let him write more code than papers ;)<br> <br></div><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><div dir="ltr"><div class="gmail_extra"><div class="gmail_quote"><div><br></div><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><div bgcolor="#FFFFFF" text="#000000">Stef<br>
    <div>Le 19/10/15 16:10, Eliot Miranda a
      écrit :<br>
    </div>
    <blockquote type="cite">
      <pre> </pre>
      <br>
      <fieldset></fieldset>
      <br>
      
      <div>Hi Robert,<br>
        <br>
        <span style="background-color:rgba(255,255,255,0)">_,,,^..^,,,_
          (phone)</span></div>
      <div><br>
        On Oct 19, 2015, at 5:03 AM, Robert Withers &lt;<a href="mailto:robert.w.withers@gmail.com" target="_blank"></a><a href="mailto:robert.w.withers@gmail.com" target="_blank">robert.w.withers@gmail.com</a>&gt;
        wrote:<br>
        <br>
      </div>
      <blockquote type="cite">
        <div><span>Hi Esteban,</span><br>
          <span></span><br>
          <span></span><br>
          <span>On 10/19/2015 05:10 AM, Esteban Lorenzano wrote:</span><br>
          <blockquote type="cite"><span></span><br>
          </blockquote>
          <blockquote type="cite"><span>Hi,</span><br>
          </blockquote>
          <blockquote type="cite"><span></span><br>
          </blockquote>
          <blockquote type="cite"><span>just to be clear.</span><br>
          </blockquote>
          <blockquote type="cite"><span>When we talk about MTVM, we talk
              about a MT-FFI, *not* a MTVM in general.</span><br>
          </blockquote>
          <blockquote type="cite"><span>In general, a “common” approach
              to MT cannot be applied in Pharo (or Smalltalk in general)
              and to get a VM *and* an image working properly is an
              effort that makes what I called massive some mails above
              like a small stone compared to a mountain.</span><br>
          </blockquote>
          <span></span><br>
          <span>Could you please help me by talking further about the
            different models and scopes of what is meant by MT?</span><br>
          <span></span><br>
          <span>a) MT-FFI I believe gives the developer a way to call
            and be invoked on callback, asynchronously. Is it so?</span><br>
        </div>
      </blockquote>
      <div><br>
      </div>
      It&#39;s a bit more than that.  It is the sharing of the VM between
      different threads, but only allowing one thread to own the VM at
      any one time, changing ownership in call out or Smalltalk process
      switch time.  This approach provides interleaved concurrency but
      not parallelism in your Smalltalk code and it means the Smalltalk
      class library doesn&#39;t have to be made thread-safe, which as
      Esteban said is a huge task.
      <div><br>
      </div>
      <div>See
        <div><a href="http://lists.gforge.inria.fr/pipermail/pharo-project/2011-January/038943.html" target="_blank">http://lists.gforge.inria.fr/pipermail/pharo-project/2011-January/038943.html</a></div>
        <div><br>
        </div>
        <div>and google &quot;eliot Miranda Simmons own thread&quot; to find more
          messages.</div>
        <div><br>
        </div>
        <div><br>
          <blockquote type="cite">
            <div><span>b) General MTVM means other system services are
                threaded, like I/O events and scheduling and heartbeat.</span><br>
            </div>
          </blockquote>
          <div><br>
          </div>
          No; at least not in my opinion. In the standard
          single-threaded VM the heartbeat is ideally a thread (it can
          be an interval timer, but that&#39;s problematic; system calls get
          interrupted), and maybe an incremental global GC could be in
          its own thread.</div>
        <div><br>
        </div>
        <div>So I&#39;m defining the MTVM to be the sharing of the VM
          between threads, and /not/ just the use of threads to
          implement non-Smalltalk sub tasks of the VM, and /not/ a
          full-blown multithreaded Smalltalk VM providing concurrent
          execution of Smalltalk processes in parallel.</div>
        <div><br>
          <blockquote type="cite">
            <div><span>I think that the right model (my
                stack/priQueue/pool intuition?) will change a Herculean
                task into a fairly straightfoward task and achievable.
                Change the problem, to get better answers.</span><br>
            </div>
          </blockquote>
          <div><br>
          </div>
          This has been well thought through and discussed.  The
          definition above is very useful.  It provides a system that
          can inter operate with concurrent code without having to
          implement a system that provides parallelism.  It is used in
          David Simmons&#39; VMs for S# etc and a similar (but less
          performant) scheme is available in Python VMs.</div>
        <div><br>
        </div>
        <div>Please, let&#39;s get this scheme working first.  I&#39;m not at
          all happy (read, extremely unhappy) that there is not much
          focus on working together to get our current VM to an
          excellent state and instead lots of work on other VMs that is
          speculative and a long way away from being production ready.
          We have a huge amount of work to do on Cog:</div>
        <div><br>
        </div>
        <div>- event-driven VM (that hence costs 0% processor time at
          idle)</div>
        <div>- 64-bits (x64 and ARM and...?)</div>
        <div>- Sista adaptive optimizer</div>
        <div>- FFI via dynamic generation of marshaling code, as
          required for efficient and correct call outs on x64</div>
        <div><span style="background-color:rgba(255,255,255,0)">-
            MTVM as defined above</span></div>
        <div><span style="background-color:rgba(255,255,255,0)">-
            an incremental global mark-sweep GC for Spur</span></div>
        <div>- running on Xen/Unikernels/containers</div>
        <div>- providing a JavaScript plugin to proved rendering and
          events so we can run an efficient VM in a web browser</div>
        <div>- a port of the Interpreter/Context VM to Spur</div>
        <div><br>
        </div>
        <div>IMO, things that can /and should/ wait are</div>
        <div>- throwing away Slang and providing a true
          written-in-pure-Smalltalk VM that is self-bootstrapped a la
          Gerardo Richarte and Xavier Burroni</div>
        <div>- a truly parallel multi/threaded VM</div>
        <div><br>
        </div>
        <div>and things we shouldn&#39;t go anywhere near are</div>
        <div>- using libffi</div>
        <div>- targeting JavaScript, Java or any other dynamic language
          de jour that happens to run in a web browser but either
          provides abysmal performance or doesn&#39;t support full Smalltalk
          semantics</div>
        <div>- implementing the VM in other VM frameworks such as PyPy
          which simply strengthens that community and weakens our own</div>
        <div><br>
        </div>
        <div>Right now there are only a handful of people who make
          commits to the VM and three who are &quot;full time&quot;, and we&#39;re all
          overloaded.  But the VM is the base of the pillar and if we
          want to provide high-quality solutions that people will pay
          money to use we have to have a high-quality VM.  In Spur we
          have a VM that is significantly faster that VW, and very
          reliable.  In Sista we will have a system that is much faster
          and can be improved upon for years to come and a system that
          can migrate to future VMs (because it is mostly Smalltalk),
          and useful support for a high quality FFI.  People like have
          stepped up and made significant contributions to give us what
          is a respectable VM that is on an arc to providing a really
          high-quality production Smalltalk VM written in Smalltalk
          produced by a very small community.  But it is now 2015 and
          Cog started 7 years ago. All the work on other VMs, deployment
          platforms etc, IMO dilutes and delays in delivering to our
          community a truly world-class VM that we can compete with
          against Java HotSpot, node.js v8, lua luajit, factor, swift et
          al.  Please get on board. We&#39;d love the help and we can
          guarantee you&#39;ll have fun and you can guarantee you&#39;ll have an
          impact.</div>
        <div><br>
          <blockquote type="cite">
            <div><span></span><br>
              <span>I appreciate you and this MT discussion.</span><br>
              <span></span><br>
              <span></span><br>
              <blockquote type="cite"><span>Said that:</span><br>
              </blockquote>
              <blockquote type="cite"><span></span><br>
              </blockquote>
              <blockquote type="cite"><span>- What is in plans is
                  MT-FFI, and that will be available eventually.</span><br>
              </blockquote>
              <blockquote type="cite"><span>- There is an approach I
                  want to re-work, that would allow us profit of
                  multicores without going multithread: the “hydra”
                  experiment made some years ago by Igor creates a good
                  basis to this. But is also a lot of of work (but a lot
                  less than a complete MT), and not a real priority for
                  now… I hope to resume work on that area some day… just
                  not anytime soon.</span><br>
              </blockquote>
              <span></span><br>
              <span>Yes, please. I recall those discussions. Hydra is
                cosmological.</span><br>
              <span></span><br>
              <span>Regards,</span><br>
              <span>Robert</span><br>
              <span></span><br>
              <blockquote type="cite"><span></span><br>
              </blockquote>
              <blockquote type="cite"><span>Esteban</span><br>
              </blockquote>
              <blockquote type="cite"><span></span><br>
              </blockquote>
              <blockquote type="cite">
                <blockquote type="cite"><span>On 18 Oct 2015, at 17:56,
                    Ben Coman &lt;<a href="mailto:btc@openinworld.com" target="_blank">btc@openInWorld.com</a>&gt;
                    wrote:</span><br>
                </blockquote>
              </blockquote>
              <blockquote type="cite">
                <blockquote type="cite"><span></span><br>
                </blockquote>
              </blockquote>
              <blockquote type="cite">
                <blockquote type="cite"><span></span><br>
                </blockquote>
              </blockquote>
              <blockquote type="cite">
                <blockquote type="cite"><span>On Sat, Oct 17, 2015 at
                    2:25 AM, Robert Withers</span><br>
                </blockquote>
              </blockquote>
              <blockquote type="cite">
                <blockquote type="cite"><span>&lt;<a href="mailto:robert.w.withers@gmail.com" target="_blank"></a><a href="mailto:robert.w.withers@gmail.com" target="_blank">robert.w.withers@gmail.com</a>&gt;
                    wrote:</span><br>
                </blockquote>
              </blockquote>
              <blockquote type="cite">
                <blockquote type="cite">
                  <blockquote type="cite"><span>Yes, exactly. I do
                      realize I was consciously changing that effort</span><br>
                  </blockquote>
                </blockquote>
              </blockquote>
              <blockquote type="cite">
                <blockquote type="cite">
                  <blockquote type="cite"><span>synchronization order.</span><br>
                  </blockquote>
                </blockquote>
              </blockquote>
              <blockquote type="cite">
                <blockquote type="cite"><span></span><br>
                </blockquote>
              </blockquote>
              <blockquote type="cite">
                <blockquote type="cite"><span>I see 64-bit being higher
                    priority than multi-threaded for the wider</span><br>
                </blockquote>
              </blockquote>
              <blockquote type="cite">
                <blockquote type="cite"><span>community.  Dealing with
                    larger in-Image data opens the door to more</span><br>
                </blockquote>
              </blockquote>
              <blockquote type="cite">
                <blockquote type="cite"><span>corporate project/funding
                    opportunities. Also simplifying the install</span><br>
                </blockquote>
              </blockquote>
              <blockquote type="cite">
                <blockquote type="cite"><span>on modern Linux platforms
                    without requiring additional 386 libraries</span><br>
                </blockquote>
              </blockquote>
              <blockquote type="cite">
                <blockquote type="cite"><span>will help acceptance
                    there.</span><br>
                </blockquote>
              </blockquote>
              <blockquote type="cite">
                <blockquote type="cite"><span></span><br>
                </blockquote>
              </blockquote>
              <blockquote type="cite">
                <blockquote type="cite">
                  <blockquote type="cite"><span>It is my humble opinion,
                      without really knowing, that 64-bit would have to
                      be redone after the MTVM completes.</span><br>
                  </blockquote>
                </blockquote>
              </blockquote>
              <blockquote type="cite">
                <blockquote type="cite"><span></span><br>
                </blockquote>
              </blockquote>
              <blockquote type="cite">
                <blockquote type="cite"><span>I would assume it was the
                    other way around. Presuming that Eliot has</span><br>
                </blockquote>
              </blockquote>
              <blockquote type="cite">
                <blockquote type="cite"><span>sponsors influencing his
                    priorities, it seems given that 64-bits will</span><br>
                </blockquote>
              </blockquote>
              <blockquote type="cite">
                <blockquote type="cite"><span>happen first.   I would
                    guess any MTVM development on the old vm would</span><br>
                </blockquote>
              </blockquote>
              <blockquote type="cite">
                <blockquote type="cite"><span>then need to be reworked.</span><br>
                </blockquote>
              </blockquote>
              <blockquote type="cite">
                <blockquote type="cite"><span></span><br>
                </blockquote>
              </blockquote>
              <blockquote type="cite">
                <blockquote type="cite">
                  <blockquote type="cite"><span>I was doing so with the
                      idea in mind that I and others</span><br>
                  </blockquote>
                </blockquote>
              </blockquote>
              <blockquote type="cite">
                <blockquote type="cite">
                  <blockquote type="cite"><span>might dig into working
                      on the VM, for threading support, while Eliot</span><br>
                  </blockquote>
                </blockquote>
              </blockquote>
              <blockquote type="cite">
                <blockquote type="cite">
                  <blockquote type="cite"><span>maintains focus on
                      64-bits...a tall order, I know.</span><br>
                  </blockquote>
                </blockquote>
              </blockquote>
              <blockquote type="cite">
                <blockquote type="cite"><span></span><br>
                </blockquote>
              </blockquote>
              <blockquote type="cite">
                <blockquote type="cite"><span>The usual downside of
                    splitting resources applies.  There are not that</span><br>
                </blockquote>
              </blockquote>
              <blockquote type="cite">
                <blockquote type="cite"><span>many &quot;others&quot; and maybe
                    they would be drawn away from helping with the</span><br>
                </blockquote>
              </blockquote>
              <blockquote type="cite">
                <blockquote type="cite"><span>64-bit vm.  If the 64-bit
                    vm goes slower for lack of resources then</span><br>
                </blockquote>
              </blockquote>
              <blockquote type="cite">
                <blockquote type="cite"><span>your footing for MTVM will
                    shifting for a longer time.  You may</span><br>
                </blockquote>
              </blockquote>
              <blockquote type="cite">
                <blockquote type="cite"><span>ultimately get where you
                    want to go faster by helping with the 64-bit</span><br>
                </blockquote>
              </blockquote>
              <blockquote type="cite">
                <blockquote type="cite"><span>vm.  The rapport built
                    with other vm devs from working on 64-bit might</span><br>
                </blockquote>
              </blockquote>
              <blockquote type="cite">
                <blockquote type="cite"><span>could then be applied to
                    MTVM.  (Of course, its your free time, so you</span><br>
                </blockquote>
              </blockquote>
              <blockquote type="cite">
                <blockquote type="cite"><span>should pursue what
                    interests you.)</span><br>
                </blockquote>
              </blockquote>
              <blockquote type="cite">
                <blockquote type="cite"><span></span><br>
                </blockquote>
              </blockquote>
              <blockquote type="cite">
                <blockquote type="cite">
                  <blockquote type="cite"><span>I was barely familiar
                      with the VM, slang, interpreter, it years ago...</span><br>
                  </blockquote>
                </blockquote>
              </blockquote>
              <blockquote type="cite">
                <blockquote type="cite">
                  <blockquote type="cite"><span>I&#39;m totally unfamiliar
                      with cog.</span><br>
                  </blockquote>
                </blockquote>
              </blockquote>
              <blockquote type="cite">
                <blockquote type="cite"><span></span><br>
                </blockquote>
              </blockquote>
              <blockquote type="cite">
                <blockquote type="cite"><span>The experience you gain
                    from working beside Esteban and Eliot on</span><br>
                </blockquote>
              </blockquote>
              <blockquote type="cite">
                <blockquote type="cite"><span>64-bit Cog/Spur could then
                    be applied to a MTVM.</span><br>
                </blockquote>
              </blockquote>
              <blockquote type="cite">
                <blockquote type="cite"><span></span><br>
                </blockquote>
              </blockquote>
              <blockquote type="cite">
                <blockquote type="cite"><span>btw, you may find these
                    threads interesting...</span><br>
                </blockquote>
              </blockquote>
              <blockquote type="cite">
                <blockquote type="cite"><span>* <a href="http://lists.pharo.org/pipermail/pharo-dev_lists.pharo.org/2015-April/108648.html" target="_blank"></a><a href="http://lists.pharo.org/pipermail/pharo-dev_lists.pharo.org/2015-April/108648.html" target="_blank">http://lists.pharo.org/pipermail/pharo-dev_lists.pharo.org/2015-April/108648.html</a></span><br>
                </blockquote>
              </blockquote>
              <blockquote type="cite">
                <blockquote type="cite"><span>* <a href="http://forum.world.st/Copy-on-write-for-a-multithreaded-VM-td4837905.html" target="_blank"></a><a href="http://forum.world.st/Copy-on-write-for-a-multithreaded-VM-td4837905.html" target="_blank">http://forum.world.st/Copy-on-write-for-a-multithreaded-VM-td4837905.html</a></span><br>
                </blockquote>
              </blockquote>
              <blockquote type="cite">
                <blockquote type="cite"><span></span><br>
                </blockquote>
              </blockquote>
              <blockquote type="cite">
                <blockquote type="cite"><span>cheers -ben</span><br>
                </blockquote>
              </blockquote>
              <blockquote type="cite">
                <blockquote type="cite"><span></span><br>
                </blockquote>
              </blockquote>
              <blockquote type="cite">
                <blockquote type="cite">
                  <blockquote type="cite"><span>I believe another item
                      on that list ought to be modernizing slang. So</span><br>
                  </blockquote>
                </blockquote>
              </blockquote>
              <blockquote type="cite">
                <blockquote type="cite">
                  <blockquote type="cite"><span>many big items!</span><br>
                  </blockquote>
                </blockquote>
              </blockquote>
              <blockquote type="cite">
                <blockquote type="cite">
                  <blockquote type="cite"><span></span><br>
                  </blockquote>
                </blockquote>
              </blockquote>
              <blockquote type="cite">
                <blockquote type="cite">
                  <blockquote type="cite"><span>Robert</span><br>
                  </blockquote>
                </blockquote>
              </blockquote>
              <blockquote type="cite">
                <blockquote type="cite">
                  <blockquote type="cite"><span></span><br>
                  </blockquote>
                </blockquote>
              </blockquote>
              <blockquote type="cite">
                <blockquote type="cite">
                  <blockquote type="cite"><span></span><br>
                  </blockquote>
                </blockquote>
              </blockquote>
              <blockquote type="cite">
                <blockquote type="cite">
                  <blockquote type="cite"><span></span><br>
                  </blockquote>
                </blockquote>
              </blockquote>
              <blockquote type="cite">
                <blockquote type="cite">
                  <blockquote type="cite"><span>On 10/16/2015 12:48 PM,
                      Stephan Eggermont wrote:</span><br>
                  </blockquote>
                </blockquote>
              </blockquote>
              <blockquote type="cite">
                <blockquote type="cite">
                  <blockquote type="cite">
                    <blockquote type="cite"><span></span><br>
                    </blockquote>
                  </blockquote>
                </blockquote>
              </blockquote>
              <blockquote type="cite">
                <blockquote type="cite">
                  <blockquote type="cite">
                    <blockquote type="cite"><span></span><br>
                    </blockquote>
                  </blockquote>
                </blockquote>
              </blockquote>
              <blockquote type="cite">
                <blockquote type="cite">
                  <blockquote type="cite">
                    <blockquote type="cite"><span>On 16-10-15 14:05,
                        Robert Withers wrote:</span><br>
                    </blockquote>
                  </blockquote>
                </blockquote>
              </blockquote>
              <blockquote type="cite">
                <blockquote type="cite">
                  <blockquote type="cite">
                    <blockquote type="cite">
                      <blockquote type="cite"><span></span><br>
                      </blockquote>
                    </blockquote>
                  </blockquote>
                </blockquote>
              </blockquote>
              <blockquote type="cite">
                <blockquote type="cite">
                  <blockquote type="cite">
                    <blockquote type="cite">
                      <blockquote type="cite"><span>Because of that
                          assumption I&#39;ve made and without the
                          responsibilities</span><br>
                      </blockquote>
                    </blockquote>
                  </blockquote>
                </blockquote>
              </blockquote>
              <blockquote type="cite">
                <blockquote type="cite">
                  <blockquote type="cite">
                    <blockquote type="cite">
                      <blockquote type="cite"><span>you have, Esteban,
                          but recognizing modernizing NB to FFI, my
                          desired</span><br>
                      </blockquote>
                    </blockquote>
                  </blockquote>
                </blockquote>
              </blockquote>
              <blockquote type="cite">
                <blockquote type="cite">
                  <blockquote type="cite">
                    <blockquote type="cite">
                      <blockquote type="cite"><span>list is:</span><br>
                      </blockquote>
                    </blockquote>
                  </blockquote>
                </blockquote>
              </blockquote>
              <blockquote type="cite">
                <blockquote type="cite">
                  <blockquote type="cite">
                    <blockquote type="cite"><span></span><br>
                    </blockquote>
                  </blockquote>
                </blockquote>
              </blockquote>
              <blockquote type="cite">
                <blockquote type="cite">
                  <blockquote type="cite">
                    <blockquote type="cite"><span></span><br>
                    </blockquote>
                  </blockquote>
                </blockquote>
              </blockquote>
              <blockquote type="cite">
                <blockquote type="cite">
                  <blockquote type="cite">
                    <blockquote type="cite"><span>I would expect the
                        least total effort to be needed by keeping the
                        work</span><br>
                    </blockquote>
                  </blockquote>
                </blockquote>
              </blockquote>
              <blockquote type="cite">
                <blockquote type="cite">
                  <blockquote type="cite">
                    <blockquote type="cite"><span>of Esteban and Eliot
                        as much as possible aligned. That is what
                        Esteban&#39;s</span><br>
                    </blockquote>
                  </blockquote>
                </blockquote>
              </blockquote>
              <blockquote type="cite">
                <blockquote type="cite">
                  <blockquote type="cite">
                    <blockquote type="cite"><span>list achieves.</span><br>
                    </blockquote>
                  </blockquote>
                </blockquote>
              </blockquote>
              <blockquote type="cite">
                <blockquote type="cite">
                  <blockquote type="cite">
                    <blockquote type="cite"><span></span><br>
                    </blockquote>
                  </blockquote>
                </blockquote>
              </blockquote>
              <blockquote type="cite">
                <blockquote type="cite">
                  <blockquote type="cite">
                    <blockquote type="cite"><span>Stephan</span><br>
                    </blockquote>
                  </blockquote>
                </blockquote>
              </blockquote>
              <blockquote type="cite">
                <blockquote type="cite">
                  <blockquote type="cite">
                    <blockquote type="cite"><span></span><br>
                    </blockquote>
                  </blockquote>
                </blockquote>
              </blockquote>
              <blockquote type="cite">
                <blockquote type="cite">
                  <blockquote type="cite">
                    <blockquote type="cite"><span></span><br>
                    </blockquote>
                  </blockquote>
                </blockquote>
              </blockquote>
              <blockquote type="cite">
                <blockquote type="cite">
                  <blockquote type="cite"><span></span><br>
                  </blockquote>
                </blockquote>
              </blockquote>
              <blockquote type="cite"><span></span><br>
              </blockquote>
            </div>
          </blockquote>
        </div>
      </div>
    </blockquote>
    <br>
  </div>

<br></blockquote></div><br><br clear="all"><div><br></div>-- <br><div><div dir="ltr"><div><span style="font-size:small;border-collapse:separate"><div>_,,,^..^,,,_<br></div><div>best, Eliot</div></span></div></div></div>
</div></div>
<br></blockquote></div><br></div></div>