The problem with RAC is the exponential decay as a function of time.  It is
just like real life: It may take most of a lifetime to become a VP, and
seconds to get fired if you screw up, or one of your subordinates does it
for you.  Likewise, if the user's computer has a problem and it does not
submit any results for a day or so, the slow dribble of pending credits
being resolved will lower the RAC considerably.

> -----Original Message-----
> From: boinc_dev [mailto:[email protected]] On Behalf
> Of McLeod, John
> Sent: Thursday, June 12, 2014 9:24 AM
> To: Richard Haselgrove; Eric J Korpela; David Anderson
> Cc: [email protected]; Josef W. Segur
> Subject: Re: [boinc_dev] EXIT_TIME_LIMIT_EXCEEDED (sorry, yes me again,
> but please read)
> 
> The reason RAC was not used was because of the delay in granting
> credit.  However, if we remember that RAC is long term, then it makes
> more sense.  If we use RAC as the estimate, then we should also have
> the client attempt to contact each of the servers for an update of RAC
> occasionally, but it would not have to be very often - once a week or
> so would probably suffice.  If it had been more than a week since the
> last connection, it would be time to try again.
> 
> -----Original Message-----
> From: boinc_dev [mailto:[email protected]] On Behalf
> Of Richard Haselgrove
> Sent: Wednesday, June 11, 2014 4:23 PM
> To: Eric J Korpela; David Anderson
> Cc: [email protected]; Josef W. Segur
> Subject: Re: [boinc_dev] EXIT_TIME_LIMIT_EXCEEDED (sorry, yes me again,
> but please read)
> 
> That one made it as far as the planning document, but no further:
> 
> http://boinc.berkeley.edu/trac/wiki/ClientSchedOctTen#Proposal:credit-
> drivenscheduling
> 
> 
> The surrogate, REC, is essentially speed * time, or back to square one.
> 
> 
> 
> >________________________________
> > From: Eric J Korpela <[email protected]>
> >To: David Anderson <[email protected]>
> >Cc: Richard Haselgrove <[email protected]>; Josef W. Segur
> <[email protected]>; "[email protected]"
> <[email protected]>
> >Sent: Wednesday, 11 June 2014, 21:03
> >Subject: Re: [boinc_dev] EXIT_TIME_LIMIT_EXCEEDED (sorry, yes me
> again, but please read)
> >
> >
> >
> >Another possibility that came to me years ago would be to use RAC
> rather than estimated duration to compute the resource allocation on
> the client side.  That way on a machine running two projects with equal
> resource share would end up spending more time running the one with
> lower granted credit per unit work.  That would encourage projects not
> to over grant (they would lose resources) or under grant (they would
> lose volunteers).
> >
> >
> >
> >
> >On Tue, Jun 10, 2014 at 1:03 PM, Eric J Korpela
> <[email protected]> wrote:
> >
> >
> >>I haven't thought about it in a while.   I had come up with a stable
> system that would but it wasn't simple and it also required projects to
> voluntarily participate.  Therefore it wouldn't have worked.
> >>
> >>The only thought I've had recently is to have a "calibration" plan
> class that has a non-SIMD non-threaded unoptimized CPU-only app_version
> that gets sent out once out of every N (~100,000) results. This (as the
> least efficient app_version) could set the pfc_scale.  Again, it would
> require project participation, so it wouldn't work.
> >>
> >>So I spend most of my time trying not to think about it.
> >>
> >>
> >>
> >>
> >>
> >>On Tue, Jun 10, 2014 at 12:12 PM, David Anderson
> <[email protected]> wrote:
> >>
> >>Are you saying we're taking the wrong approach?
> >>>Any other suggestions?
> >>>
> >>>
> >>>On 10-Jun-2014 11:51 AM, Eric J Korpela wrote:
> >>>
> >>> >For credit purposes, the standard is peak FLOPS,
> >>>> >i.e. we give credit for what the device could do,
> >>>> >rather than what it actually did.
> >>>> >Among other things, this encourages projects to develop more
> efficient apps.
> >>>>
> >>>>It does the opposite because many projects care more about
> attracting volunteers
> >>>>than they do about efficient computation.
> >>>>
> >>>>First: Per second of run time,  a host gets the same credit for a
> non-optimized
> >>>>stock app as it does for an optimized stock app.  There's no
> benefit to the
> >>>>volunteer to go to a project with optimized apps.  In fact there's
> a benefit for
> >>>>users to compile an optimized app for use at a non-optimized
> project where their
> >>>>credit will be higher.  Every time we optimize SETI@home we get
> bombarded by users
> >>>>of non-stock optimized apps get angry because their RAC goes down.
>  That makes it a
> >>>>disincentive to optimize.
> >>>>
> >>>>Second:  This method encourages projects to create separate apps
> for GPUs rather
> >>>>than separate app_versions.  Because GPUs obtain nowhere near their
> advertised rates
> >>>>for real code, a separate GPU app can earn 20 to 100x the credit of
> a GPU
> >>>>app_version of an app that also has CPU app_versions.
> >>>>
> >>>>Third: It encourages projects to not use the BOINC credit granting
> mechanisms.  To
> >>>>compete with projects that have GPU only apps, some projects grant
> outrageous credit
> >>>>for everything.
> >>>>
> >>>>
> >>>>
> >>>>
> >>>>
> >>>>On Tue, Jun 10, 2014 at 11:34 AM, David Anderson
> <[email protected]
> >>>>
> >>>><mailto:[email protected]>> wrote:
> >>>>
> >>>>    For credit purposes, the standard is peak FLOPS,
> >>>>    i.e. we give credit for what the device could do,
> >>>>    rather than what it actually did.
> >>>>    Among other things, this encourages projects to develop more
> efficient apps.
> >>>>
> >>>>    Currently we're not measuring this well for x86 CPUs,
> >>>>    since our Whetstone benchmark isn't optimized.
> >>>>    Ideally the BOINC client should include variants for the most
> common
> >>>>    CPU features, as we do for ARM.
> >>>>
> >>>>    -- D
> >>>>
> >>>>
> >>>>    On 10-Jun-2014 2:09 AM, Richard Haselgrove wrote:
> >>>>
> >>>>        Before anybody leaps into making any changes on the basis
> of that observation, I
> >>>>        think we ought to pause and consider why we have a
> benchmark, and what we
> >>>>        use it for.
> >>>>
> >>>>        I'd suggest that in an ideal world, we would be measuring
> the actual running
> >>>>        speed
> >>>>        of (each project's) science applications on that particular
> host,
> >>>>        optimisations and
> >>>>        all. We gradually do this through the runtime averages
> anyway, but it's hard to
> >>>>        gather a priori data on a new host.
> >>>>
> >>>>        Instead of (initially) measuring science application
> performance, we measure
> >>>>        hardware performance as a surrogate. We now have (at least)
> three ways of
> >>>>        doing that:
> >>>>
> >>>>        x86: minimum, most conservative, estimate, no optimisations
> allowed for.
> >>>>        Android: allows for optimised hardware pathways with vfp or
> neon, but
> >>>>        doesn't relate
> >>>>        back to science app capability.
> >>>>        GPU: maximum theoretical 'peak flops', calculated from card
> parameters, then
> >>>>        scaled
> >>>>        back by rule of thumb.
> >>>>
> >>>>        Maybe we should standardise on just one standard?
> >>>>
> >>>>
> >>>>
>         ------------------------------__------------------------------
> __------------------------
> >>>>
> >>>>             *From:* Richard Haselgrove
> <[email protected]
> >>>>
>         <mailto:[email protected]>>
> >>>>
> >>>>             *To:* Josef W. Segur <[email protected]
> >>>>
>         <mailto:[email protected]>>; David Anderson
> >>>>             <[email protected]
> <mailto:[email protected]>>
> >>>>             *Cc:* "[email protected]
> <mailto:[email protected]>"
> >>>>        <[email protected]
> <mailto:[email protected]>>
> >>>>
> >>>>             *Sent:* Tuesday, 10 June 2014, 9:37
> >>>>             *Subject:* Re: [boinc_dev] EXIT_TIME_LIMIT_EXCEEDED
> (sorry, yes me
> >>>>        again, but
> >>>>
> >>>>             please read)
> >>>>
> >>>>
>         http://boinc.berkeley.edu/__gitweb/?p=boinc-
> v2.git;a=__commit;h=__7b2ca9e787a204f2a57f390bc7249b__b7f9997fea
> >>>>        <http://boinc.berkeley.edu/gitweb/?p=boinc-
> v2.git;a=commit;h=7b2ca9e787a204f2a57f390bc7249bb7f9997fea>
> >>>>
> >>>>              >__________________________________
> >>>>
> >>>>              > From: Josef W. Segur <[email protected]
> >>>>
>         <mailto:[email protected]> <mailto:[email protected]
> >>>>
> >>>>        <mailto:[email protected]>>>
> >>>>              >To: David Anderson <[email protected]
> >>>>
>         <mailto:[email protected]> <mailto:[email protected]
> >>>>        <mailto:[email protected]>__>>
> >>>>              >Cc: "[email protected]
> <mailto:[email protected]>
> >>>>        <mailto:boinc_dev@ssl.__berkeley.edu
> <mailto:[email protected]>>"
> >>>>             <[email protected]
> <mailto:[email protected]>
> >>>>        <mailto:boinc_dev@ssl.__berkeley.edu
> <mailto:[email protected]>>>;
> >>>>
> >>>>        Eric J Korpela
> >>>>             <[email protected]
> <mailto:[email protected]>
> >>>>
>         <mailto:[email protected].__edu
> <mailto:[email protected]>>>;
> >>>>
> >>>>        Richard Haselgrove
> >>>>             <[email protected]
> <mailto:[email protected]>
> >>>>
>         <mailto:r.haselgrove@__btopenworld.com
> <mailto:[email protected]>>>
> >>>>
> >>>>
> >>>>              >Sent: Tuesday, 10 June 2014, 2:19
> >>>>              >Subject: Re: [boinc_dev] EXIT_TIME_LIMIT_EXCEEDED
> (sorry, yes me
> >>>>        again, but
> >>>>             please read)
> >>>>              >
> >>>>              >
> >>>>              >Consider Richard's observation:
> >>>>              >
> >>>>              >>>     It appears that the Android Whetstone
> benchmark used in the BOINC
> >>>>             client has
> >>>>              >>>     separate code paths for ARM, vfp, and NEON
> processors: a vfp
> >>>>        or NEON
> >>>>             processor
> >>>>              >>>     will report that it is significantly faster
> than a
> >>>>        plain-vanilla ARM.
> >>>>              >
> >>>>              >If that is so, it distinctly differs from the x86
> Whetstone which
> >>>>        never uses
> >>>>             SIMD, and is truly conservative as you would want for
> 3).
> >>>>              >--
> >>>>              >                               Joe
> >>>>              >
> >>>>              >
> >>>>              >
> >>>>              >On Mon, 09 Jun 2014 16:43:17 -0400, David Anderson
> >>>>        <[email protected] <mailto:[email protected]>
> >>>>
> >>>>             <mailto:[email protected]
> <mailto:[email protected]>__>> wrote:
> >>>>              >
> >>>>              >> Eric:
> >>>>              >>
> >>>>              >> Yes, I suspect that's what's going on.
> >>>>              >> Currently the logic for estimating job runtime
> >>>>              >> (estimate_flops() in sched_version.cpp) is
> >>>>              >> 1) if this (host, app version) has > 10 results,
> use (host, app
> >>>>        version)
> >>>>             statistics
> >>>>              >> 2) if this app version has > 100 results, use app
> version statistics
> >>>>              >> 3) else use a conservative estimate based on
> p_fpops.
> >>>>              >>
> >>>>              >> I'm not sure we should be doing 2) at all,
> >>>>              >> since as you point out the first x100 or 1000
> results for an app
> >>>>        version
> >>>>              >> will generally be from the fastest devices
> >>>>              >> (and even in the steady state,
> >>>>              >> app version statistics disproportionately reflect
> fast devices).
> >>>>              >>
> >>>>              >> I'll make this change.
> >>>>              >>
> >>>>              >> -- David
> >>>>              >>
> >>>>              >> On 09-Jun-2014 8:10 AM, Eric J Korpela wrote:
> >>>>              >>> I also don't have direct access to the server as
> well, so I'm
> >>>>        mostly guessing.
> >>>>              >>> Having separate benchmarks for neon and VFP means
> there's a broad
> >>>>        bimodal
> >>>>              >>> distribution for the benchmark results.  Where
> the mean falls
> >>>>        depends upon
> >>>>             the mix
> >>>>              >>> of machines.  In general the neon machines (being
> newer and
> >>>>        faster) will report
> >>>>              >>> first and more often, so early on the PFC
> distribution will
> >>>>        reflect the fast
> >>>>              >>> machines.  Slower machines will be underweighted.
>  So the work will be
> >>>>             estimated to
> >>>>              >>> complete quickly, and some machines will time
> out.  In SETI beta, it
> >>>>             resolves itself
> >>>>              >>> in a few weeks.  I can't guarantee that it will
> anywhere else.
> >>>>              >>>
> >>>>              >>> We see this with every release of a GPU app.  The
> real
> >>>>        capabilities of graphics
> >>>>              >>> cards vary by orders of magnitude from the
> estimate and by more
> >>>>        from each
> >>>>             other.
> >>>>              >>> The fast cards report first and most every else
> hits days of timeouts.
> >>>>              >>>
> >>>>              >>> One possible fix so to increase the timeout
> limits for the first 10
> >>>>             workunits for a
> >>>>              >>> host_app_version, until host based estimates take
> over.
> >>>>              >>>
> >>>>              >>>
> >>>>              >>>
> >>>>              >>>
> >>>>              >>> On Mon, Jun 9, 2014 at 2:02 AM, Richard
> Haselgrove
> >>>>             <[email protected]
> <mailto:[email protected]>
> >>>>
>         <mailto:r.haselgrove@__btopenworld.com
> <mailto:[email protected]>>
> >>>>              >>> <mailto:r.haselgrove@__btopenworld.com
> >>>>        <mailto:[email protected]>
> >>>>
> >>>>             <mailto:r.haselgrove@__btopenworld.com
> >>>>
> >>>>        <mailto:[email protected]>>>> wrote:
> >>>>              >>>
> >>>>              >>>     I think Eric Korpela would be the best person
> to answer that
> >>>>        question,
> >>>>             but I
> >>>>              >>>     suspect 'probably not': further investigation
> over the weekend
> >>>>        suggests
> >>>>             that the
> >>>>              >>>     circumstances may be SIMAP-specific.
> >>>>              >>>
> >>>>              >>>     It appears that the Android Whetstone
> benchmark used in the BOINC
> >>>>             client has
> >>>>              >>>     separate code paths for ARM, vfp, and NEON
> processors: a vfp
> >>>>        or NEON
> >>>>             processor
> >>>>              >>>     will report that it is significantly faster
> than a
> >>>>        plain-vanilla ARM.
> >>>>              >>>
> >>>>              >>>     However, SIMAP have only deployed a single
> Android app, which I'm
> >>>>             assuming only
> >>>>              >>>     uses ARM functions: devices with vfp or NEON
> SIMD vectorisation
> >>>>             available would
> >>>>              >>>     run the non-optimised application much slower
> than BOINC expects.
> >>>>              >>>
> >>>>              >>>     At my suggestion, Thomas Rattei (SIMAP
> admistrator) increased the
> >>>>              >>>     rsc_fpops_bound multiplier to 10x on Sunday
> afternoon. I note
> >>>>        that the
> >>>>             maximum
> >>>>              >>>     runtime displayed on
> >>>>
>         http://boincsimap.org/__boincsimap/server_status.php
> >>>>
> >>>>        <http://boincsimap.org/boincsimap/server_status.php> has
> >>>>              >>>     already increased from 11 hours to 14 hours
> since he did that.
> >>>>              >>>
> >>>>              >>>     Thomas has told me "We've seen that
> [EXIT_TIME_LIMIT_EXCEEDED]
> >>>>        a lot.
> >>>>             However,
> >>>>              >>>     due to Samsung PowerSleep, we thought these
> are mainly "lazy"
> >>>>        users
> >>>>             just not
> >>>>              >>>     using their phone regularly for computing."
> He's going to
> >>>>        monitor how this
> >>>>              >>>     progresses during the remainder of the
> current batch, and I've
> >>>>        asked
> >>>>             him to keep
> >>>>              >>>     us updated on his observations.
> >>>>              >>>
> >>>>              >>>
> >>>>              >>>
> >>>>
>               >>>      >__________________________________
> >>>>
> >>>>              >>>      > From: David Anderson
> <[email protected]
> >>>>        <mailto:[email protected]>
> >>>>
>              <mailto:[email protected]
> <mailto:[email protected]>__>
> >>>>        <mailto:[email protected]
> <mailto:[email protected]>
> >>>>
> >>>>             <mailto:[email protected]
> <mailto:[email protected]>__>>>
> >>>>              >>>      >To: [email protected]
> >>>>        <mailto:[email protected]>
> <mailto:boinc_dev@ssl.__berkeley.edu
> >>>>        <mailto:[email protected]>>
> >>>>             <mailto:boinc_dev@ssl.__berkeley.edu
> >>>>        <mailto:[email protected]>
> <mailto:boinc_dev@ssl.__berkeley.edu
> >>>>
> >>>>        <mailto:[email protected]>>>
> >>>>
> >>>>              >>>     >Sent: Monday, 9 June 2014, 3:48
> >>>>              >>>      >Subject: Re: [boinc_dev]
> EXIT_TIME_LIMIT_EXCEEDED (sorry, yes me
> >>>>             again, but
> >>>>              >>>     please read)
> >>>>              >>>      >
> >>>>              >>>      >
> >>>>              >>>      >Does this problem occur on SETI@home?
> >>>>              >>>      >-- David
> >>>>              >>>      >
> >>>>              >>>      >On 07-Jun-2014 2:51 AM, Richard Haselgrove
> wrote:
> >>>>              >>>      >
> >>>>              >>>      >> 2) Android runtime estimates
> >>>>              >>>      >>
> >>>>              >>>      >> The example here is from SIMAP. During a
> recent pause between
> >>>>             batches, I noticed
> >>>>              >>>      >> that some of my 'pending validation'
> tasks were being slow
> >>>>        to clear:
> >>>>              >>>      >>
> >>>>
>         http://boincsimap.org/__boincsimap/results.php?hostid=__349248
> >>>>
> >>>>
> <http://boincsimap.org/boincsimap/results.php?hostid=349248>
> >>>>              >>>      >>
> >>>>              >>>      >> The clearest example is the third of
> those three workunits:
> >>>>              >>>      >>
> >>>>
>         http://boincsimap.org/__boincsimap/workunit.php?wuid=__57169928
> >>>>
> >>>>
> <http://boincsimap.org/boincsimap/workunit.php?wuid=57169928>
> >>>>              >>>      >>
> >>>>              >>>      >> Four of the seven replications have
> failed with 'Error while
> >>>>             computing', and
> >>>>              >>>      >> every one of those four is an
> EXIT_TIME_LIMIT_EXCEEDED on an
> >>>>             Android device.
> >>>>              >>>      >>
> >>>>              >>>      >> Three of the four hosts have never
> returned a valid result
> >>>>        (total
> >>>>             credit zero),
> >>>>              >>>      >> so they have never had a chance to
> establish an APR for
> >>>>        use in runtime
> >>>>              >>>      >> estimation: runtime estimates and bounds
> must have been
> >>>>        generated
> >>>>             by the server.
> >>>>              >>>      >>
> >>>>              >>>      >> It seems - from these results, and others
> I've found
> >>>>        pending on
> >>>>             other machines -
> >>>>              >>>      >> that SIMAP tasks on Android are aborted
> with
> >>>>             EXIT_TIME_LIMIT_EXCEEDED after ~6
> >>>>              >>>      >> hours elapsed. For the new batch released
> today, SIMAP are
> >>>>        using a
> >>>>             3x bound
> >>>>              >>>      >> (which may be a bit low under the
> circumstances):
> >>>>              >>>      >>
> >>>>
>               >>>      >>
> <rsc_fpops_est>13500000000000.__000000</rsc_fpops_est>
> >>>>              >>>     >>
> <rsc_fpops_bound>__40500000000000.000000</rsc___fpops_bound>
> >>>>
> >>>>              >>>      >>
> >>>>              >>>      >> so I deduce that the tasks when first
> issued had a runtime
> >>>>        estimate
> >>>>             of ~2 hours.
> >>>>              >>>      >>
> >>>>              >>>      >> My own tasks, on a fast Intel i5
> 'Haswell' CPU (APR 7.34
> >>>>        GFLOPS),
> >>>>             take over half
> >>>>              >>>      >> an hour to complete: two hours for an ARM
> device sounds
> >>>>             suspiciously low. The
> >>>>              >>>      >> only one of my Android wingmates to have
> registered an APR
> >>>>              >>>      >>
> >>>>
> >>>>
> 
> (http://boincsimap.org/__boincsimap/host_app_versions.__php?hostid=7710
> 33
> >>>>
> <http://boincsimap.org/boincsimap/host_app_versions.php?hostid=771033>)
> is
> >>>>
> >>>>              >>>     showing
> >>>>              >>>      >> 1.69 GFLOPS, but I have no way of knowing
> whether that APR was
> >>>>             established
> >>>>              >>>     before
> >>>>              >>>      >> or after the task in question errored
> out.
> >>>>              >>>      >>
> >>>>              >>>      >> From experience - borne out by current
> tests at
> >>>>        Albert@Home, where
> >>>>             server logs
> >>>>              >>>      >> are helpfully exposed to the public -
> initial server
> >>>>        estimates can
> >>>>             be hopelessly
> >>>>              >>>      >> over-optimistic. These two are for the
> same machine:
> >>>>              >>>      >>
> >>>>              >>>      >> 2014-06-04 20:28:09.8459 [PID=26529]
> [version] [AV#716]
> >>>>             (BRP4G-cuda32-nv301)
> >>>>              >>>      >> adjusting projected flops based on PFC
> avg: 2124.60G
> >>>>        2014-06-07
> >>>>             09:30:56.1506
> >>>>              >>>      >> [PID=10808] [version] [AV#716] (BRP4G-
> cuda32-nv301) setting
> >>>>             projected flops
> >>>>              >>>     based
> >>>>              >>>      >> on host elapsed time avg: 23.71G
> >>>>              >>>      >>
> >>>>              >>>      >> Since SIMAP have recently announced that
> they are leaving
> >>>>        the BOINC
> >>>>             platform at
> >>>>              >>>      >> the end of the year (despite being an
> Android launch
> >>>>        partner with
> >>>>             Samsung), I
> >>>>              >>>      >> doubt they'll want to put much effort
> into researching
> >>>>        this issue.
> >>>>              >>>      >>
> >>>>              >>>      >> But if other projects experimenting with
> Android
> >>>>        applications are
> >>>>             experiencing a
> >>>>              >>>      >> high task failure rate, they might like
> to check whether
> >>>>              >>>     EXIT_TIME_LIMIT_EXCEEDED
> >>>>              >>>      >> is a significant factor in those
> failures, and if so,
> >>>>        consider the
> >>>>             other
> >>>>              >>>      >> remediation approaches (apart from
> outliers, which isn't
> >>>>        relevant
> >>>>             in this case)
> >>>>              >>>      >> that I suggested to Eric Mcintosh at LHC.
> >>>>              >
> >>>>              >
> >>>>              >
> >>>>
>              _________________________________________________
> >>>>
> >>>>             boinc_dev mailing list
> >>>>        [email protected]
> <mailto:[email protected]>
> >>>>
>         <mailto:boinc_dev@ssl.__berkeley.edu
> <mailto:[email protected]>>
> >>>>
> >>>>        http://lists.ssl.berkeley.edu/__mailman/listinfo/boinc_dev
> >>>>
> >>>>        <http://lists.ssl.berkeley.edu/mailman/listinfo/boinc_dev>
> >>>>             To unsubscribe, visit the above URL and
> >>>>             (near bottom of page) enter your email address.
> >>>>
> >>>>
> >>>>
>     _________________________________________________
> >>>>
> >>>>    boinc_dev mailing list
> >>>>    [email protected] <mailto:[email protected]>
> >>>>
>     http://lists.ssl.berkeley.edu/__mailman/listinfo/boinc_dev
> >>>>
> >>>>    <http://lists.ssl.berkeley.edu/mailman/listinfo/boinc_dev>
> >>>>    To unsubscribe, visit the above URL and
> >>>>    (near bottom of page) enter your email address.
> >>>>
> >>>>
> >>>>
> >>
> >
> >
> >
> _______________________________________________
> boinc_dev mailing list
> [email protected]
> http://lists.ssl.berkeley.edu/mailman/listinfo/boinc_dev
> To unsubscribe, visit the above URL and
> (near bottom of page) enter your email address.
> _______________________________________________
> boinc_dev mailing list
> [email protected]
> http://lists.ssl.berkeley.edu/mailman/listinfo/boinc_dev
> To unsubscribe, visit the above URL and
> (near bottom of page) enter your email address.

_______________________________________________
boinc_dev mailing list
[email protected]
http://lists.ssl.berkeley.edu/mailman/listinfo/boinc_dev
To unsubscribe, visit the above URL and
(near bottom of page) enter your email address.

Reply via email to