On Wed, Oct 24, 2012 at 12:52:03PM -0700, Ross Walker wrote:
>
> >It is not a problem of gmail's spam filter, it is the best I
> >have seen. Even if messages are marked as spam they are
> >accessible from Spam folder and go away only after certain
> >amount of time. About the real issue at the end.
>
> Agreed but in my experience it gets way too many false positives and if
> one is access mail remotely via another client the messages in junk are
> not delivered as part of a POP3 connection for example hence they can go
> undetected for a long time.
True, but I think that it would be reasonable to expect that in
case of problems users would check it. In my case I'm using it
also trough IMAP, so I can jump to other folders and check.
>
> >>The problem is two fold. Firstly gmail can greylist things so message
> >> delivery for gmail accounts can often be substantially delayed and
> >>second
> >> gmail is I believe hiding your own copy of the message from you since it
> >> is trying to be too smart for it's own good and detecting the same
> >>message
> >> already in your sent items.
> >
> >First time I hear that gmail might greylisting mails, never
> >happened to me in 5 years. Any noticable delay might be only
> >because using email client which checks for mail periodically,
> >so they show up in the program only after next check or Google
> >is having unconfined server problems (doesn't happen often).
>
> post log
> Oct 24 11:30:34 2012 (3322) post to amber from gam4rei.gmail.com,
> size=6179, message-id=<20121024183025.GE27961.RD-HC.apollo.lv>, success
>
> smtp log
> Oct 24 11:30:34 2012 (3322) <20121024183025.GE27961.RD-HC.apollo.lv> smtp
> to amber for 1372 recips, completed in 0.257 seconds
>
> exim main log
> 2012-10-24 11:30:33 1TR5ir-0000JB-Oj <= gam4rei.gmail.com
> H=mail-lb0-f171.google.com [209.85.217.171] P=esmtps X=TLSv1:RC4-SHA:128 S
> =6108 id=20121024183025.GE27961.RD-HC.apollo.lv T="Re: [AMBER] [PATCH 4/8]
> sqm: Fix OpenMP build"
> ..
> ..
> 2012-10-24 11:30:39 1TR5is-0000JG-NU aspmx.l.google.com
> [2607:f8b0:400e:c03::1b] Network is unreachable
> 2012-10-24 11:30:39 1TR5is-0000JG-NU aspmx.l.google.com
> [2607:f8b0:400e:c03::1b] Network is unreachable
> ..
>
> ..
> 2012-10-24 11:30:48 1TR5is-0000JG-Rj SMTP error from remote mail server
> after RCPT TO:<.. ...googlemail.com>: host gmail-smtp-in.l.google.com
> [74.125.141.26]: 451-4.3.0 Multiple destination domains per transaction is
> unsupported. Please\n451 4.3.0 try again. o4si22306281paw.25
> ..
>
> 2012-10-24 11:30:56 1TR5is-0000JG-Rj SMTP error from remote mail server
> after RCPT TO:<.. ...googlemail.com>: host gmail
> -smtp-in.l.google.com [74.125.141.26]: 451-4.3.0 Multiple destination
> domains per transaction is unsupported. Please\n451 4.3.0 try
> again. f9si22268238paz.203
> ..
>
> 2012-10-24 11:32:51 1TR5is-0000JG-Rj == .. ...gmail.com R=dnslookup
> T=remote_smtp defer (-53): retry time not reached for any host
>
> ..
>
> Looks a lot like grey listing to me. Perhaps semantically it is different
> but either way it leads to a delay in delivery of mail to some gmail
> addresses.
It seems that the issue is with Google's policy (or bug) which
results in denied mail delivery and required retries and thus
some delays for batch mails which are sent to multiple domains.
According to this thread it started from this spring:
https://productforums.google.com/forum/?fromgroups=#!topic/gmail/qQ3vDWouukk
I haven't read all the referenced threads there, but it seems
that a lot of people are pointing at Google, but it at ISPs so it
is not clear whose issue it really is.
Delay itself is too small for me to really notice, my POP mail
client checks it every 10 min so I wouldn't notice 5 min delay.
But you are right, it seems to be there.
> >>
> >> Simple solution is to use a non-gmail account which would actually be
> >> better for everyone on the list. People should not be using 'anonymous'
> >> accounts to post to what is a scientific discussion list. I am tempted
> >>to
> >> enforce a rule that everyone posting either has an account where we can
> >> identify who they are OR signs their email with a signature identifying
> >> their full name, institution and affiliation.
> >
> >No, that would be a wrong way to go. First, institution mail has
> >no place for high volume mailing lists. Second, people's names
> >and affiliations has nothing to do with scientific discussions.
> >Third, we are not talking here as representatives of our
> >institutions.
>
> Actually I beg to differ here. Anonymity has no place science and indeed
> one has no right to it when expressing scientific opinions which is the
> main basis of this mailing list. To promote discussion of science and
> scientific problems as they center around molecular dynamics and the use
> of the AMBER molecular dynamics research software. If one chooses in a
> scientific community to make statements or offer opinions behind the
> shroud of anonymity then such opinions should be viewed with extreme
> skepticism. I fully believe that if one wants respect and recognition in
> science, which indeed is what scientific reputation is based on then they
> should be prepared to stand by any statements they make right or wrong.
> This has been the basis of scientific discussion for hundreds of years and
> one should not casually dismiss that on the basis that it is now
> substantially easier to make comments 'pseudo' anonymously than it used to
> be.
Then this is a difference in opinions on what this list is
about. I'm viewing it as a technical support forums for Amber
and AmberTools, where people can get advice about the usage of
said tools, but this is not a place to design your experiments
or construct theories. Indeed, for general science discussions
the website points to chemistry.ccl.net. And I would be really
surprised if people would expect to be cited from any mailing
list.
Sure anonymous comment should be scrutinized, but the same can
be said for any other scientific comment. If one accepts opinion
just because it is expressed by a professor it can also lead
astray. Especially since mailing lists tend to have a really
diverse public, the language barriers and scientific or even
cultural background can lead to surprising conclusions. And one
really needs to work all the details out on his own to be able
to really understand the subject. At least it is so for me and I
often question the validity of propositions if the connection is
not really clear to me.
>
> As for talking as a representative of your institution, I'm really sorry
> to inform you but unfortunately what you are posting is a reflection of
> your institutional affiliation. As a graduate student, postdoc or faculty
> at a university your voices and opinions as they relate to your work at
> said university are a reflection on the university. Indeed, almost all
> students and staff have signed a document signing over their work to
> ownership by the university. If you are not at a university in some
> capacity and/or what you are posting here is in no way related to your
> research at the university then you are correct you are not representing
> the research group you belong to or your institution. However, if what you
> post here relates to your research, or was undertaken during your expected
> hours of employment or study then unfortunately you do represent the
> institution whether you like it or not.
In that sense I'm representing also humanity just by being one
of us and doing what humans do. And I agree that my posts are
reflecting on my university since I'm a part of it. But when I
said that I'm not speaking as a representative of my institution
I meant that my posts are not expressing opinions of my
university, they are solely my personal views, and thus my
affiliation should not be mandated just to participate in these
discussions. If that makes my word weaker than anyone else's who
have stated their affiliation, then so be it, everyone has to
decide on their own if anything what I'm suggesting is any good.
>
> Relating to my first point about anonymity. Including your details and
> institution with your comments also adds significant weight to the value
> of such comments since people viewing them will base their opinion on both
> your reputation and that of the institution you are at. This is why
> research papers have full affiliations listed on them.
It will add only so much weight as much one believes in
reputation and how familiar one is with said person and the
named institution. For people without that familiarity much more
weight would be in the role each participant has in the
community and the area of activity. For example to me a statement
from someone saying "your patches are wrong they will not be
taken, because ..." would mean a lot more if the person
would state that he is a developer working on component Z.
> >
> >You are already moderating signups for amber-developers which is
> >hilarious for publicly readable list. And you have absolutely no
> >pointer on website for where to send user contributions for
> >AmberTools (Bug tracker? Maintainer mail? Source repository with
> >the relevant information around? Give something!). I can tell
> >that my little experiment to send these patches was really
> >frustrating exactly because of these two reasons.
>
> The moderation of the developers list has nothing to do with posts being
> public, it is purely so we have some semi-autonomous control over who
> receives emails and announcements sent to that list. It is a pretty simple
> matter to subscribe to the list. Simply subscribe, email the list owner
> (me) with your affiliation to the AMBER developers or just a very simple
> explanation of why you want to receive messages and announcements related
> to the development of the code and I'll happily approve the subscription.
Then please add a description to amber-developers subscription
page or website, that one has to provide said information in
subscription request. And if I ever will have a need to be
subscribed there I will send such a mail. It was not obvious for
me until it was pointed out, the page states to send a blank
mail.
>
> As for AmberTools. We have a full bug tracking system that we use
> internally and we can add people to that who are happy to use it properly,
> appreciating that people's time is very limited.
Is there a strong reason why the bug tracker is not public? I
would really like to know about known bugs in software I'm using
so I can be careful what I'm doing. For that even any write
rights are not necessary.
> We also have full git
> repositories for the code. Given the research nature of the software we
> tend to be very wary of accepting user contributions that are sent
> directly. It is extremely easy to break such fragile codes and it is also
> very easy to write code that gives the wrong answer scientifically.
> Community scientific code without control on what gets added can be VERY
> dangerous. Thus for almost any scientific code it tends to be better for
> contributions to go through the maintainer of that code. The simplest way
> to do that is to post to the AMBER mailing list and ask who maintains
> various pieces of code.
Central place to reach all of them would have been nice. I
thought that amber-developers was such a place, but apparently
you can't just subscribe to it the same way as to this list.
> The website says:
>
> "We hope to add new functionality to AmberTools as additional programs
> become available. If you have suggestions for what might be added, please
> contact us."
Yes, and I managed to do it with second attempt. The case for
authors of particular codes is true, but it is not really clear
who maintains the build system and I wanted to reach everyone in
one go, also because I wanted feedback (or at least reach) all
of the people since these changes touch the glue between many
components.
>
> Which seems pretty self explanatory to me. The authors of the various
> packages are all stated, none of us contribute anonymously, although we
> may be extremely busy writing grant proposals and reports these days which
> means it can take a few weeks to get replies but it shouldn't be difficult
> to contact one of us and then if we are not the right person we'll refer
> you to the right person.
>
> In your case a post to the AMBER list asking the best way to make
> contributions and explaining what you found would certainly have been the
> best starting point. Then be patient and people will contact you. I don't
> want to discourage people making contributions. Far from it, we value such
> things, it is just that these days, what with current funding climates,
> people tend to be extremely overworked and so it is critical to be patient
> in such matters.
Understood, I was too impatient, but having worked on it for a
whole weekend I wanted to get it out there quickly.
> >>
> >> A more complex solution would be to check your gmail settings and see if
> >> there is a way to prevent this sort of thing happening but I don't know
> >>of
> >> a solution. If you find one please let me know and I can add it to the
> >> list subscription instructions.
> >
> >The issue seems to be with gmail as you already suggest. The
> >thing is that gmail recognizes duplicates using message IDs and
> >removes them without providing a configuration option to change
> >this behaviour.
> >
> >So the solution would be to set mailman to change message ID (if
> >possible) or just document that this particular option will not
> >work with gmail (or any other account automatically and
> >unconditionally removing mails with the same IDs). The messages
> >are still saved (as sent) so they are available, but without any
> >modification mailman does to bounced mail.
>
> If you know how to do this, or other recommended mailman settings I'd be
> grateful to hear suggestions. I took a look through the various settings
> and couldn't see anything immediately obvious for how to have the message
> sent with a different message id.
I really don't know anything about mailman configuration, but at
their list they have discussed it at length:
http://www.mail-archive.com/mailman-developers.python.org/msg06753.html
I haven't read the thread, so I don't know what was the
conclusion and whether the issue has been raised since. But it
seems to me that this is something were it would be really hard
or impossible to get everyone happy.
So I think the best course of action would be just to add a note
on subscription page or in configuration page if possible, that
gmail (or provider who doesn't allow duplicate messages) users
will not get their own mail bounced back to them, but they can
still see it in Sent.
>
> >It would also help if the archive would be updated as soon as
> >there are new posts.
>
> It takes over 45 minutes right now to process the mail archive and update
> the page. Hence why I have it set to do it just once every few days. It
> has been on my radar for a long time to improve this, have it update the
> page incrementally rather than reprocessing the entire mail spool, but it
> simply hasn't managed to float to the top of my todo list. If you know a
> simple way to do it please let me know. I know mailman has an archive
> system built in but it is not very flexible and I couldn't see an easy way
> of importing 13 years of archived messages into it.
Oh, sounds like a major undertaking. I have just used SourceForge
and their archive is updating on every message or so, but they
probably have a custom system and the scale of the project is
completely different. Have you considered something like Gmane
for this list?
http://gmane.org/
Best Regards,
Reinis
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Wed Oct 24 2012 - 16:00:04 PDT
