mailman
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Mailman's Approval Duties Denied Me.


From: Bob Proulx
Subject: Re: Mailman's Approval Duties Denied Me.
Date: Sat, 24 Feb 2018 13:55:46 -0700
User-agent: Mutt/1.9.3 (2018-01-21)

Hi Ralph,

> I've now done those, and everything in
> http://listhelper.nongnu.org/mailmanconf.html given by Karl, for all
> three nmh-* lists, and cleared out any lists of automatically discarded,
> etc., so I start from scratch.

:-)

> I noticed address@hidden is subscribed to nmh-workers.  Is that
> related to their helping with list moderation, or just coincidence with
> them liking nmh?  :-)

It is to keep the Bayes engine trained.  How do you keep the Bayes
database trained on what is the normal accepted email to a mailing
list?  Easy.  Subscribe it to the mailing list.  It then trains
continuously and automatically on whatever is the accepted mail
flowing through the mailing lists.

All of the mailing list mail shares a common Bayes database.  That is
sub-optimal but it was easy to set up and splitting them into 1500+
databases was harder.  Therefore assuming that all mailing list mail
is similar then it works pretty well.  Except for that assumption.  If
the Bayes were split for each mailing list then training on accepted
messages would be much more important.

> > I don't see any funky characters that would prevent the grep from
> > working.
> >
> >   address@hidden:/u/listhelper/www/moderate$ grep nmh- fsf-lists | od -Ax 
> > -tx1z -v -c
> 
> Thanks, that has me (re-)learning od's `z'!  I'd have done
> 
>     ... | tr -d ' -~' | od -c
> 
> >   000000  6e  6d  68  2d  61  6e  6e  6f  75  6e  63  65  0a  6e  6d  68  
> > >nmh-announce.nmh<
> >            n   m   h   -   a   n   n   o   u   n   c   e  \n   n   m   h
> >   000010  2d  63  6f  6d  6d  69  74  73  0a  6e  6d  68  2d  77  6f  72  
> > >-commits.nmh-wor<
> >            -   c   o   m   m   I   t   s  \n   n   m   h   -   w   o   r
> >   000020  6b  65  72  73  0a                                              
> > >kers.<
> >            k   e   r   s  \n
> 
> The `i' of `commits' above is uppercase in the -c output, but lowercase,
> 0x69, in the -tx1z, and in the earlier grep.  Not that it matters since
> we're talking nmh-announce.

Oh, good catch!  However that is an emacs abbreviation that was
triggered inadvertantly when I pasted in the block of text.  I type in
a lower case 'i' and emacs converts it as an abreviation expansion
into an upper case 'I' for me automatically.  A personal customization
that I added because I was so slack about hitting the shift key in the
right time to catch the single letter 'I' character.  I usually look
for that when I am pasting them into my email windows but sometimes I
forget and I forgot to look at that time.  It is lower case in the
file and that can be seen because the hex value was listed as 0x69
right above it in the line. :-)

> >   Last updated 2018-02-23 04:14:21 UTC
> >   bug-wget (1)
> >   freetype-devel (1)
> >   gnash-dev (2)
> >   qemu-devel (1)
> >   savannah-hackers (1)
> >   summer-of-code (1)
> >   debbugs-submit
> >
> > Where each of those names are links off to Mailman's moderator page.
> > Here is the first one.
> >
> >   https://lists.gnu.org/mailman/admindb/bug-wget
> 
> Ah, OK, seems foolproof.  :-)  I suppose a listhelper could have their
> own script that scans the output of that admindb URL and all the others,
> bypassing fsf-lists.

Well, yes, and, ahem, *I* happen to be one of those.  I grab the raw
data and do a text console interface so that I can run through a block
of them all at once and not have to log into each of them individually
through the web interface.  But I update it each time with the current
set of fsf-lists and the other lists and so the result should be the
same.

And as far as I can tell I am the only one doing something like that
at this time.  I originally wrote the script in Ruby.  Ruby is a
delightful language.  But I made the mistake of using the ruby
libraries supplied by the community.  They worked great for a
release.  But then after an update the libraries became unstable for
processing this type of data with regards to unicode handling.  The
email contains all kinds of invalid data.  It's spam!  So you can
imagine that it would be invalid data.  The Ruby libs started to throw
invalid character exceptions from deep inside.  I spent several days
trying to chase down the underlying problems so that I could fix them
but most of that code is "meta" generated code.  It doesn't really
exist and is created on the fly.  Which really made it impossible for
me to get to the root of the issue.  I term the type of code I found
in the community libraries write-only code.  It was good at the time
for the original author but impossible for me to tinker out.

And at that point the program became too unstable for other people to
use.  I stil use it.  For about 80% of the email it works okay.  And
for the 20% that it backtraces on (effectively the script equivalent
of a core dump) I fall back to handling from the web interface.

One of these days I will get some time and will rewrite it again in a
different language that has higher quality libraries for the
processing.  Personally I would fall back to using Perl since it has
always had excellent quality in the CPAN libraries.  I have perl
scripts that have been doing complex tasks for years and years over
many updates and they have not had their APIs thrashed.  But who knows
this might be something I would write in Go-lang or something in order
to have a task to learn a new language.  If I want to learn a new
language I must have a task to do in it or there is lack of motivation
to continue with it.

> Yep, that's fine.  I'm happy there's nothing more that can be done.

Sorry!

Bob



reply via email to

[Prev in Thread] Current Thread [Next in Thread]